Citation
Transitive Inference in Temporal Lobe Epilepsy

Material Information

Title:
Transitive Inference in Temporal Lobe Epilepsy
Creator:
Barker, Marie
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (77 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Psychology
Clinical and Health Psychology
Committee Chair:
Bauer, Russell M.
Committee Members:
Marsiske, Michael
Fennell, Eileen B.
Loring, David
Roper, Steven N.
Graduation Date:
8/8/2009

Subjects

Subjects / Keywords:
Demography ( jstor )
Epilepsy ( jstor )
Hippocampus ( jstor )
Inference ( jstor )
Learning ( jstor )
Memory ( jstor )
Paradigms ( jstor )
Seizures ( jstor )
Temporal lobe ( jstor )
Temporal lobe epilepsy ( jstor )
Clinical and Health Psychology -- Dissertations, Academic -- UF
association, epilepsy, hippocampal, hippocampus, inference, lateralization, lobectomy, memory, relational, temporal, transitive
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Psychology thesis, Ph.D.

Notes

Abstract:
Recent findings in cognitive neuroscience reveal that transitive inference (TI) tasks, which require the formation and recognition of stimulus associations across experiences, have good specificity in the measurement of hippocampal functioning. Extant research has focused on animals and healthy adults. This study is the first to apply the TI paradigm in temporal lobe epilepsy (TLE), which is a syndrome that provides a model to study hippocampal contributions to memory. Primary aims were 1) to examine TI performance and relationship to side of surgery in TLE and 2) to compare the clinical utility of the TI task to standard neuropsychological tests. Participants included 24 patients with TLE, who had undergone anterior temporal lobectomy (ATL; left n=8, right n=16), and 24 healthy controls. They completed a computer-based TI task, which was adapted from a paradigm that has demonstrated selective right hippocampal activation in functional imaging studies (Heckers et al., 2004). During training, participants view pairs of patterned shapes and learn the 'winner' in each pair (e.g., A > B, B > C). They are tested on their ability to recollect the correct response for previously seen pairs and to make inferences about novel pairings (e.g., A > C). The critical condition involves making inferences across a series of overlapping pairs that form a hierarchy (A > B > C > D > E). On the test, patients who had undergone right ATL performed significantly worse than healthy controls on TI for visual information. Left ATL patients performed in the intermediate range; however, the task did not discriminate between patients based on side of surgery. Results provide some evidence of a laterality effect and suggest that TI may be sensitive to hippocampally-mediated memory function. There is a clear need for better neuropsychological measures to assess language non-dominant (usually right) temporal lobe function, given the poor sensitivity and specificity of current tests. In this study, the conventional nonverbal memory measure, the Rey Complex Figure Test, did not discriminate between groups. The TI task and the conventional nonverbal measure yielded similar operating characteristics with good positive predictive power but poor sensitivity. The TI task showed modest clinical promise, and modifications that may improve its clinical utility are suggested. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2009.
Local:
Adviser: Bauer, Russell M.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31
Statement of Responsibility:
by Marie Barker.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Barker, Marie. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
8/31/2010
Classification:
LD1780 2009 ( lcc )

Downloads

This item has the following downloads:


Full Text





Training performance on overlapping pairs was also evaluated (Figure 3-6). The main

effect of training was significant [F(2, 90) = 25.74, p < .001, r12 = .36]. Bonferroni adjusted post-

hoc tests showed that reaction times improved across training from Block 1 to Block 2 (p < .001)

and from Block 1 to Block 3 (p < .001). Neither the main effect of group [F(2, 45) = 3.03, p =

.06, 12 .12], nor the interaction between group and training block [F(4, 90) = 1.14, p = .34, r2

.05] was significant. Given the trend toward group differences showing right ATL performing

worse than controls and the limited power (.56) to detect group differences, exploratory one-way

ANOVAs were conducted to examine reaction times for each block. A trend toward group

differences was observed on Block 1 [F(2, 45) = 2.92, p = .06, r2 = .11] and Block 3 [F(2, 45) =

2.91,p = .06,1= .11].

In summary, participants demonstrated quicker reaction times as training progressed for

both the non-overlapping and the overlapping sets. On the final training block for non-

overlapping pairs, controls showed faster reaction times than either of the two epilepsy groups.

No significant group differences were observed on reaction times during training for the

overlapping pairs.

Reaction Time: Test

Test performance was evaluated using a 3 (Group: right ATL, left ATL, control) x 4

(Condition: Non-overlapping Trained, Non-overlapping Inference, Overlapping Trained,

Overlapping Inference) mixed between-within ANOVA. The main effect of group was not

significant, F(2, 43) = 2.01,p = .15, 12 = .08, (Figure 3-7). There was a main effect of condition,

F(3, 129) = 6.90, p < .001, r2 = .14. Bonferroni adjusted post-hoc tests showed that participants

exhibited significantly slower reaction times on the Overlapping Trained condition compared to

the Non-overlapping Trained (p = .001) and Non-overlapping Inference (p = .006) conditions.

The main effect was moderated by a significant interaction between group and condition, F(6,









clinical assessments incorporate measures of figural reproduction, such as the Visual

Reproductions subtest of the Wechsler Memory Scale-Third Edition (Wechsler, 1997b) or the

Rey Complex Figure Test (Meyers & Meyers, 1995), which ask the examinee to draw the figure

from memory after a single presentation. A few studies have demonstrated some lateralizing

value of figural reproduction tests in TLE (Glosser, Cole, Khatri, DellaPietra, & Kaplan, 2002;

Jones-Gotman, 1991).

However, the sensitivity of these tests to detect memory dysfunction specific to right TLE

is questionable. For instance, a meta-analysis was conducted on 33 studies of TLE that used the

Wechsler Memory Scale subtests of Logical Memory and Visual Reproductions to assess verbal

and nonverbal memory respectively (Lee, Yip, and Jones-Gotman, 2002). While the verbal task

was sensitive to left hemisphere dysfunction both pre- and post-operatively, the efficacy of the

nonverbal task to assess right hemisphere dysfunction was not confirmed.

Further compelling evidence of the poor lateralizing capability of figural reproduction tests

comes from a study by Barr and colleagues (1997). In a sample of approximately 750 patients,

they detected no significant differences between patients with right and left TLE on either

Wechsler Memory Scale Visual Reproductions or the Rey Complex Figure Test. Their analyses

were well powered and controlled for potential confounds, limiting any concerns about

inadequacies in design. Similar negative findings have been noted in other studies of TLE using

the Brief Visuospatial Memory Test-Revised (Benedict, 1997), a figural reproduction test with

multiple learning trials, and the Continuous Visual Memory Test (Trahan & Larrabee, 1988), a

figural learning and recognition test (Barr, Morrison, Zaroff, & Devinsky, 2004; Snitz, Roman,

& Beniak, 1996).









Griffith, H. R., Pyzalski, R. W., O'Leary, D., Magnotta, V., Bell, B., Dow, C., et al. (2003). A
controlled quantitative MRI volumetric investigation of hippocampal contributions to
immediate and delayed memory performance. Journal of Clinical and Experimental
Neuropsychology, 25, 1117-1127.

Hannula, D. E., Tranel, D., & Cohen, N. J. (2006). The long and the short of it: Relational
memory impairments in amnesia, even at short lags. The Journal ofNeuroscience, 26,
8352-8359.

Heaton, R. K. (1981). Wisconsin Card Sorting Test (WCST). Odessa, FL: Psychological
Assessment Resources.

Heckers, S., Zalesak, M., Weiss, A. P., Ditman, T., & Titone, D. (2004). Hippocampal activation
during transitive inference in humans. Hippocampus, 14, 153-162.

Helmstaedter, C. (2004). Neuropsychological aspects of epilepsy surgery. Epilepsy & Behavior,
5(Suppl. 1), S45-55.

Helmstaedter, C., Gleissner, U., Di Perna, M., & Elger, C. E. (1997). Relational verbal memory
processing in patients with temporal lobe epilepsy. Cortex, 33, 667-678.

Helmstaedter, C., & Kurthen, M. (2001). Memory and epilepsy: Characteristics, course, and
influence of drugs and surgery. Current Opinion in Neurology, 14, 211-216.

Helmstaedter, C., Pohl, C., & Elger, C. E. (1995). Relations between verbal and nonverbal
memory performance: Evidence of confounding effects particularly in patients with right
temporal lobe epilepsy. Cortex, 31, 345-355.

Hermann, B. P., Seidenberg, M., Dow, C., Jones, J., Rutecki, P., Bhattacharya, A., et al. (2006).
Cognitive prognosis in chronic temporal lobe epilepsy. Annals ofNeurology, 60, 80-87.

Hermann, B. P., Seidenberg, M., Schoenfeld, J., & Davies, K. (1997). Neuropsychological
characteristics of the syndrome of mesial temporal lobe epilepsy. Archives of Neurology,
54, 369-376.

Jackson, O., & Schacter, D. L. (2004). Encoding activity in anterior medial temporal lobe
supports subsequent associative recognition. Neuroimage, 21, 456-462.

Jones-Gotman, M. (1991). Localization of lesions by neuropsychological testing. Epilepsia,
32(Suppl. 5), S41-52.

Kneebone, A. C., Lee, G. P., Wade, L. T., & Loring, D. W. (2007). Rey Complex Figure: figural
and spatial memory before and after temporal lobectomy for intractable epilepsy. Journal
of the International Neuropsychological Society, 13, 664-671.

Koehler, S., Danckert, S., Gati, J. S., & Menon, R. S. (2005). Novelty responses to relational and
non-relational information in the hippocampus and parahippocampal region: A comparison
based on event-related MRI. Hippocampus, 15, 763-774.









patients with medial temporal lobe damage (Kroll, Knight, Metcalfe, Wolf, & Tulving, 1996)

and in functional imaging studies (Giovanello et al., 2004; Meltzer & Constable, 2005).

Transitive inference

Another approach to test relational memory involves training subjects with distinct stimuli

that share common elements and then testing whether these experiences have been linked in

memory to solve new problems (Eichenbaum, 2000). In the TI paradigm, subjects are exposed

to a series of overlapping stimulus pairs (e.g., A>B, B>C, where ">" means "is preferred to" in

regards to the likelihood of obtaining reinforcement when that item is selected over the other). A

measure of relational memory is provided by the subject's ability to make an inference about

stimuli that were not previously presented together, but are related through the overlapping pairs.

For example, TI would be demonstrated by the knowledge that A>C. The overlapping pairs are

likely stored as a flexible representation that can be manipulated mentally to solve the transitive

problem (Heckers, Zalesak, Weiss, Ditman, & Titone, 2004). Although this effect is "episodic"

in the sense that it arises out of exposure to the stimuli, it is "inferential" in that it involves

reference to a memory representation that is not based on direct experience, since during

learning, A and C were never presented together.

The TI paradigm is based on a well-validated animal model, and the capacity for TI has

been demonstrated in rodents (Dusek & Eichenbaum, 1997), pigeons (von Fersen, Wynne,

Delius, & Staddon, 1991), and nonhuman primates (McGonigle & Chalmers, 1977) that were

trained in instrumental learning formats. This paradigm has also been applied to humans in a

comparative approach. In healthy adults, TI has primarily been tested using visual materials,

such as geometric designs (Heckers et al., 2004, Heckers & Zalesak, 2009), faces (Nagode &

Pardo, 2002), and face-house pairings (Preston, Shrager, Dudukovic, & Gabrieli, 2004).









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ..............................................................................................................4

L IST O F T A B L E S .................................................................................................. . 7

LIST OF FIGURES .................................. .. ..... ..... ................. .8

ABSTRAC T ..........................................................................................

CHAPTER

1 BACKGROUND AND SIGNIFICANCE........................................... 11

R elation al M em ory ....................... .... ...... .... .. ............ ..................... 11
The Hippocampus: The Physiological Basis for Relational Memory .............................12
Paradigm s to A ssess Relational M em ory ................................. ............. .................. 13
Paired associate learning ............................................. ...... ................13
Transitive inference ................ ... ................................ ............... 14
Converging Evidence for the Role of the Hippocampus in Relational Memory ............15
A n im al m o d els ..................................................... ................ 15
Neuroimaging of healthy adults ............. ......................................... 16
Studies of amnesia..................... ............... ...... ... ...... ............. 18
Relational Memory: Application to Temporal Lobe Epilepsy .......................... ..................19
The Role of Neuropsychological Testing in the Surgical Treatment of Epilepsy...........20
Assessment of Memory Functioning in TLE ............................. .. ..............21
Improving Detection of Memory Dysfunction in Right TLE .............. ...................23
P u rp o se of C u rrent Stu dy ........................................................................................... .. 2 5

2 M E T H O D S .......................................................................................................2 6

P articip an ts .........................................................................2 6
P ro c e d u re s .............................................................................................2 6
Recruitment Strategy ................. .. ......... ..................26
A ssessm ent P ro cedu res ............................................................................................. 2 7
M measures .............. .............. ................................................. 28
Transitive Inference Task .............. ......... ......... ........28
T ra in in g ................................................................2 8
T e st ............. .. ............... ............................. ..................... 3 0
Scoring and interpretation ......................................................... 30
Post-test assessment ............ .............. ....... .........3 1
Standard N europsychological Tests ........................................................ 32









data on the verbal measure, it was eliminated from the study. In the future, it will be important

to refine this methodology to develop a verbal TI analogue of a similar level of difficulty, in

order to facilitate comparison and to evaluate laterality effects.

Lastly, the current study evaluated TI in patients who had undergone ATL. Examining the

paradigm in post-surgical epilepsy patients is an initial step. The next step would be to evaluate

TI in pre-surgical epilepsy patients to determine whether the paradigm shows sensitivity to right

temporal dysfunction in this patient group. The clinical implications in this patient group are

more relevant, as the eventual applied goal of this research is to develop more effective pre-

surgical tools to assist in the diagnosis of localized/lateralized seizure onset and to predict the

likelihood of post-surgical decline in cognitive function (Helmstaedter, 2004).

Conclusion

The current study is the first to examine TI in TLE. The study showed that TI for visual

information is difficult for patients who underwent right ATL. Results provide some limited

support for the material-specificity of TI. More broadly, the findings suggest a link between

hippocampal function and TI, which is consistent with the relational memory concept. In light of

the lack of behavioral measures sensitive to language non-dominant hippocampal function, the

TI paradigm may have clinical promise. In future research, it would be beneficial to continue

studying TI capacities in the TLE population, using some of the modifications suggested in

preceding sections and expanding the work to include pre-surgical epilepsy patients.









Table 3-1. Demographic and clinical characteristics by group
Right ATL (n=16) Left ATL (n=8) Control (n=24)
Age 43.9(11.1) 43.9 (15.5) 39.7 (15.4)
Years of education 14.3 (2.7) 13.4 (2.1) 15.7 (2.9)
Gender (% female) 44% 50% 54%
Race (% Caucasian) 87.5% 100% 75%
History of mild-moderate TBI 50.0% 25.0% 0%
History of psychiatric diagnosis 25.0% 25.0% 0%
Handedness (laterality quotient) 0.84 (.18) 0.52 (.67) 0.86 (.16)
Age at onset of epilepsy (yrs.) 22.1 (14.4) 17.4 (13.8)
Duration of epilepsy (yrs.) 17.4 (12.3) 24.0 (16.9)
Family history of epilepsy 31.2% 50.0% -
History of febrile seizures 25.0% 12.5% -
Pre-surg. seizure frequency 8.4 (11.5) 10.1(11.4) -
(per month)
Seizure classification
Complex partial 93.7% 87.5% -
Generalized 6.2% 12.5% -
Pre-surgical MRI results
No abnormalities 37.5% 12.5% -
Hippocampal sclerosis 56.2% 62.5% -
Other lesion/ abnormality 6.2% 25.0% -
Language dominance (Wada)b
Left 100% 71% -
Mixed 0% 29% -
Months since surgery 60.1 (27.7) 44.9 (23.2) -
Post-surgical seizure freedom
Seizure free 68.7% 62.5% -
Aura only 18.7% 12.5% -
Continued seizures 12.5% 25.0% -
Note. Data presented as mean (standard deviation), except for those variables listed as percent of
sample. aIn control group, one subject was missing years of education. bOne left ATL and two
right ATL patients missing Wada language dominance data.









(i.e., false prediction of right ATL in someone with left ATL) against the rate of true positives

(i.e., correct classification of right ATL). The area under the curve represents how well each

measure predicts right-sided resection, with larger areas indicative of stronger predictors.

ROC curve analysis for the Overlapping Inference score revealed area under the curve of

.68 (SE= .12, 95% CI= .44-.91) (Figure 3-8). A cut score of 75% yielded the following

operating characteristics: sensitivity = 50%, specificity = 87.5%, positive predictive power =

89%, and negative predictive power = 47%. Thus, while a score below the cut-off indicated a

high probability that a patient had undergone right ATL, only half the patients who underwent

right ATL were detected with this cut score. Using a higher cut score identified more patients

who underwent right temporal lobe surgery, but yielded an unacceptably high rate of false

positives. For example, a cut-score of 85% yielded sensitivity = 62.5%, specificity = 62.5%,

positive predictive power = 76.9%, and negative predictive power = 45.5%.

To further examine TI performance, independent samples t-tests and Pearson chi-square

tests contrasted right ATL patients who performed above (n=8) and below (n=8) the 75% cut

score. No significant group differences (p > .10) were detected on demographic or clinical

variables, including age, years of education, age at onset of epilepsy, duration of epilepsy,

months since surgery, language or memory dominance, or presence of pre-surgical hippocampal

sclerosis. Group differences were detected on self-reported hierarchical awareness [2(1) = 7.27,

p = .007], such that 100% of patients who performed below the cut score denied hierarchical

awareness, while only 37.5% of patients who performed above the cut score did not endorse it.

ROC curve analysis was also conducted to evaluate the predictive utility of the Rey

Complex Figure Test (Figure 3-8). This analysis used the age-corrected score for Delayed

Recall. Area under the curve was .66 (SE = .11, 95% CI = .43-.88). Operating characteristics









replacement strategy was applied to improve normality of the reaction time distributions. For

each variable, scores greater than 2.5 standard deviations above the mean were replaced with the

reaction time value at 2.5 standard deviations above the mean. In subsequent analyses,

assumptions of the general linear model were tested when indicated, and appropriate corrections

were applied if the assumptions were not met.

Aim 1: Transitive Inference Task Performance

Performance on the TI task was evaluated using accuracy and latency scores for the

training and test conditions. It was hypothesized that inferences about the overlapping pairs

would be impaired in patients who underwent right ATL relative to patients with left ATL and

healthy controls. It was expected that right ATL patients may show deficits on other conditions

of the task, but deficits would be most evident on the Overlapping Inference condition.

Accuracy: Training

Descriptive statistics for accuracy (percentage correct) are presented in Table 3-2.

Training performance for each stimulus set was evaluated using a 3 (Group: right ATL, left

ATL, control) x 3 (Training: Block 1, Block 2, Block 3) mixed between-within ANOVA.

First, training for the non-overlapping pairs was examined (Figure 3-1). For the within-

subjects factor, Mauchly's test was significant (p < .05), suggesting a violation of the sphericity

assumption and indicating that the variances in the differences between blocks were not

equivalent. Thus, Greenhouse-Geisser df corrections are reported. Levene's test also reached

significance for Block 3 of Training, indicating that the assumption of equality of error variances

across groups was not met for this block. The effect of training was significant, F(1.61, 72.62) =

83.56, p < .001, r2 = .65. Bonferroni adjusted post-hoc tests showed significant improvement

across training from Block 1 to Block 2 (p < .001) and from Block 2 to Block 3 (p < .001).

Neither the main effect of group [F(2, 45) = 2.35, p = .11, rf2 = .09], nor the interaction between









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

TRANSITIVE INFERENCE IN TEMPORAL LOBE EPILEPSY

By

Marie D. Barker

August 2009

Chair: Russell M. Bauer
Major: Psychology

Recent findings in cognitive neuroscience reveal that transitive inference (TI) tasks, which

require the formation and recognition of stimulus associations across experiences, have good

specificity in the measurement of hippocampal functioning. Extant research has focused on

animals and healthy adults. This study is the first to apply the TI paradigm in temporal lobe

epilepsy (TLE), which is a syndrome that provides a model to study hippocampal contributions

to memory. Primary aims were 1) to examine TI performance and relationship to side of surgery

in TLE and 2) to compare the clinical utility of the TI task to standard neuropsychological tests.

Participants included 24 patients with TLE, who had undergone anterior temporal

lobectomy (ATL; left n=8, right n=16), and 24 healthy controls. They completed a computer-

based TI task, which was adapted from a paradigm that has demonstrated selective right

hippocampal activation in functional imaging studies (Heckers et al., 2004). During training,

participants view pairs of patterned shapes and learn the "winner" in each pair (e.g., A>B, B>C).

They are tested on their ability to recollect the correct response for previously seen pairs and to

make inferences about novel pairings (e.g., A>C). The critical condition involves making

inferences across a series of overlapping pairs that form a hierarchy (A>B>C>D>E).









CHAPTER 1
BACKGROUND AND SIGNIFICANCE

Relational Memory

Relational memory refers to the capacity to create a flexible and integrated representation

of an experience that mediates associations among elements of the experience (Cohen, Poldrack,

& Eichenbaum, 1997). Both the individual elements as well as their larger structure are encoded

in relational memory. Anatomic data suggest that the hippocampus is essential for relational

binding, linking multiple inputs together to represent their relationships and to code overlapping

features across different experiences (Cohen et al., 1999).

The concept of relational memory fits into the current framework of multiple memory

systems. This framework arose out of research on amnesia, which first implicated the

hippocampus in memory (Scoville & Milner, 1957). The current framework focuses on a

distinction between declarative and nondeclarative memory processing (Squire, 2004).

Declarative memory is characterized by the conscious recollection of facts and events, while

nondeclarative memory does not involve conscious recollection and is demonstrated through

facilitation of performance rather than recollection.

Given that relational memory involves the mediation of associations between information,

it can be integral for both declarative and nondeclarative memory. For example, relational

memory is essential for a form of declarative memory that involves representing episodes. In

episodic memory, an individual encodes how different elements relate to form a representation of

a complex event (e.g., a party, a final examination), what is unique about the event, and how it

links to other episodic memories through common elements (Eichenbaum, 2000; O'Reilly &

Rudy, 2001).









TRANSITIVE INFERENCE IN TEMPORAL LOBE EPILEPSY


By

MARIE D. BARKER


















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2009









Level of initial learning was considered as a source of variability in test performance.

Although groups achieved fairly comparable levels of learning during training, differences were

observed on several individual training blocks. Therefore, analyses of covariance (ANCOVAs)

to control for initial learning were conducted. The covariate was accuracy on the third training

block for the appropriate stimulus set. Again, group was the between-subjects factor and

accuracy score for the test condition was the dependent variable. These analyses revealed the

same pattern as reported in the previous analysis. For Non-Overlapping Inference, initial

learning was significantly related to performance, F(1, 44) = 84.16, p < .001, 12 = .66. There

was no main effect of group, F(2, 44) = 1.10, p = .34, r1 = .05. For Overlapping Inference,

initial learning was also significantly related to performance, F(1, 44) = 45.43, p < .001, r12 = .51.

A main effect of group was observed, F(2, 44) = 4.93, p = .01, r12 = .18. Bonferroni adjusted

post-hoc tests revealed that patients with right ATL performed worse than controls, t(44) = 2.99,

p = .01, on this TI condition.

Intra-group variability in test performance was observed, as evidenced by large standard

deviations (Table 3-2). Demographic and clinical variables were examined to determine whether

any were related to test performance (Table 3-3). Age was significantly related to performance

in three of the four test conditions, such that older age was associated with lower accuracy

scores. The relationship between age and test performance did not appear to be an artifact of an

underlying relationship between duration of epilepsy and test performance as these two variables

were not significantly correlated. Moreover, within the epilepsy sample, a series of partial

correlations controlling for duration of epilepsy revealed moderate relationships between age and

these test conditions (r = .39 to .50).









that are specifically designed to assess relational memory may show better sensitivity to

hippocampally-mediated memory function in TLE.

Purpose of Current Study

The primary aim of the present study was to examine TI in patients with intractable TLE

who underwent ATL. There are no published studies of TI in TLE, and results have the potential

to provide support for relational memory theory in a new population. TI performance and its

relationship to side of surgery were examined. It was hypothesized that material-specific deficits

in performance would be observed; such that TI for patterned visual stimuli would be impaired

in patients with right ATL relative to patients with left ATL and healthy controls. Deficits were

expected to be most evident on the TI condition, although it was expected that right ATL patients

may demonstrate deficits on other conditions involving premise pairs from the overlapping

hierarchy. In addition, the current study sought to replicate behavioral findings from Heckers

and colleagues' (2004) study, such as the TI effect for response latencies, which predicts that

reaction times will be slowest for pairs involving TI.

The secondary aim was to compare the clinical utility of the TI task to established clinical

memory tests. The experimental TI task addresses some of the concerns about conventional

nonverbal memory tests by utilizing a recognition format and stimuli that are difficult to

verbalize. The Rey Complex Figure Test was selected as the standard measure of nonverbal

memory because it is widely used in epilepsy surgery centers to assess memory dysfunction

associated with right TLE (Barr et al., 1997; Frank & Landeira-Fernandez, 2008). It was

hypothesized that the TI task would provide more accurate detection of right-sided resection than

the standard test of nonverbal memory.









Eichenbaum (in press) showed that hippocampal damage produced after mice learned an

overlapping sequence did not substantially affect original learning but resulted in severe

impairment in subsequent TI. This finding implies that the hippocampus is important for

accessing the representation in a way necessary to perform TI.

Neuroimaging of healthy adults

Neuroimaging studies of healthy adults also provide evidence of the role of the

hippocampus in relational memory. In a review of the functional imaging literature, Cohen and

colleagues (1999) concluded that the relational memory concept provided a better explanation of

the data than other accounts of hippocampal functioning, which focus on novelty, the explicit-

implicit memory distinction, or spatial mapping. The functional imaging studies typically rely

on a difference analysis, such that the pattern of activation observed during an item memory task

or other "control task" is compared to activation observed during a relational memory task.

Areas that are selectively activated during the relational memory task, in comparison to the

"control task," are considered to reflect a unique contribution to relational memory.

Functional imaging studies indicate that the framework for relational memory is laid down

during the encoding process. For instance, greater hippocampal activation was observed during

training for overlapping premise pairs that permitted TI compared to those that did not (Nagode

& Pardo, 2002). Similarly, other studies have demonstrated hippocampal involvement during

encoding of novel associations involving visual information (Rombouts et al., 1997; Sperling et

al., 2001). Later studies have expanded on this work to show that the extent of hippocampal

activation during relational encoding correlates with performance on subsequent memory tasks.

Hippocampal activity is greater during encoding of relational information that is successfully

remembered on subsequent recall and recognition tasks (Davachi & Wagner, 2002; Jackson &

Schacter, 2004; Staresina & Davachi, 2006).









After the ordering task, participants were asked whether they were aware that the stimuli in

the overlapping set formed a hierarchy. The proportion of participants who endorsed

hierarchical awareness was 31.2% of patients with right ATL, 42.9% of patients with left ATL,

and 62.5% of controls. A chi-square test indicated that these did not represent significant group

differences, 2(2) = 3.87, p =.14.

Reaction Time: Training

Descriptive statistics for median reaction time are presented in Table 3-4. Training

performance for each stimulus set was evaluated using a 3 (Group: right ATL, left ATL, control)

x 3 (Training: Block 1, Block 2, Block 3) mixed between-within ANOVA.

First, training on the non-overlapping stimulus set was evaluated (Figure 3-5). For the

within-subjects factor, Mauchly's test was significant (p < .05), suggesting a violation of the

sphericity assumption. Thus, Greenhouse-Geisser df corrections are reported. The effect of

training was significant [F(1.31, 59.11) = 45.50, p < .001, r2 = .50]. Bonferroni adjusted post-

hoc tests showed improved reaction times from Block 1 to Block 2 (p < .001) and from Block 2

to Block 3 (p = .02). There was also a main effect of group [F(2, 45) = 3.63, p = .03, 12 = .14];

however, Bonferroni-corrected post-hoc analyses did not show significant group differences.

Post-hoc tests that were not adjusted for multiple comparisons revealed that controls had faster

reaction times than patients with left ATL (p = .02) or right ATL (p = .05). The interaction

between group and training block was not significant [F(2.63, 59.11) = .23, p = .85, 12 = .01].

Given that the power to detect differences between groups was only .64, exploratory one-way

ANOVAs were conducted to further examine performance on each block. Group differences

were detected on Block 3 [F(2, 45) = 6.08, p = .005, 12 = .21]. Control subjects demonstrated

faster reaction times on Block 3 than patients with left ATL, t(45) = 3.01, p = .01, and patients

with right ATL, t(45) = 2.62, p = .04.









were calculated for cut scores at T = 30 (sensitivity = 40%, specificity = 100%, positive

predictive power = 100%, negative predictive power = 47%) and at T = 35 (sensitivity = 60%,

specificity = 50%, positive predictive power = 69%, negative predictive power = 40%).












1500
1400
1300
1200
1100
1000
900
800
700
600
500


Non-overlapping Non-overlapping Overlapping
trained inference trained
Test condition


SRATL
E LATL
O Control


Overlapping
inference


Figure 3-7. Reaction time by test condition.


0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
1 Specificity 1 Specificity
A B

Figure 3-8. Receiver operating characteristic (ROC) curves for the prediction of right anterior
temporal lobectomy. A) Transitive inference (accuracy for overlapping pairs). B) Rey Complex
Figure Test delayed recall.








100
95
90
9 85
80
75
70
a 65
60
55
50
Non-overlapping Non-overlapping Overlapping Overlapping
trained inference trained inference
Test condition

Figure 3-3. Accuracy by test condition.


BD
Transitive inference pairs


Figure 3-4. Accuracy for overlapping pairs requiring transitive inference.


SRATL
* LATL
o Control


E RATL
* LATL
O Control


I


I









Nagode, J. C., & Pardo, J. V. (2002). Human hippocampal activation during transitive inference.
Neuroreport, 13, 939-944.

O'Brien, C. E., Bowden, S. C., Bardenhagen, F. J., & Cook, M. J. (2003). Neuropsychological
correlates of hippocampal and rhinal cortex volumes in patients with mesial temporal
sclerosis. Hippocampus, 13, 892-904.

Oldfield, R. C. (1971). The assessment and analysis of handedness. The Edinburgh Inventory.
Neuropsychologia, 9, 97-113.

Olson, I. R., Page, K., Moore, K. S., Chatterjee, A., & Verfaellie, M. (2006). Working memory
for conjunctions relies on the medial temporal lobe. The Journal ofNeuroscience, 26,
4596-4601.

O'Reilly, R. C., & Rudy, J. W. (2001). Conjunctive representations in learning and memory:
Principles of cortical and hippocampal function. Psychological Review, 108, 311-345.

Preston, A. R., & Gabrieli, J. D. (2002). Different functions for different medial temporal lobe
structures? Learning andMemory, 9, 215-217.

Preston, A. R., Shrager, Y., Dudukovic, N. M., & Gabrieli, J. D. (2004). Hippocampal
contribution to the novel use of relational information in declarative memory.
Hippocampus, 14, 148-152.

Prince, S. E., Daselaar, S. M., & Cabeza, R. (2005). Neural correlates of relational memory:
Successful encoding and retrieval of semantic and perceptual associations. Journal of
Neuroscience, 25, 1203-1210.

Psychological Corporation. (1999). Wechsler Abbreviated Scale ofIntelligence (WASI). San
Antonio, TX: Author.

Reitan, R. M. (1958). Validity of the Trail Making Test as an indicator of organic brain damage.
Perceptual andMotor .k//l, 8, 271-276.

Rickard, T. C., & Grafman, J. (1998). Losing their configural mind: Amnesic patients fail on
transverse patterning. Journal of Cognitive Neuroscience, 10, 509-524.

Rombouts, S. A., Machielsen, W. C., Witter, M., Barkhof, F., Lindeboom, J., & Scheltens, P.
(1997). Visual association encoding activates the medial temporal lobe: A functional
magnetic resonance imaging study. Hippocampus, 7, 594-601.

Ryan, J. D., Moses, S. N., & Villate, C. (2009). Impaired relational organization of propositions,
but intact transitive inference, in aging: Implications for understanding neural integrity.
Neuropsychologia, 47, 338-353.

Savage, G. R., Saling, M. M., Davis, C. W., & Berkovic, S. F. (2002). Direct and indirect
measures of verbal relational memory following anterior temporal lobectomy.
Neuropsychologia, 40, 302-316.









BIOGRAPHICAL SKETCH

Ms. Barker grew up in Bristol, Tennessee. She graduated from the University of the South

at Sewanee in 2000 with a Bachelor of Science degree in Psychology and a minor in German.

After graduating from college, she studied psychology in Munich, Germany with a Fulbright

Fellowship. She then worked for two years as a Research Assistant at the Yale University Child

Study Center before beginning graduate school in Clinical Psychology at the University of

Florida. She earned her Master of Science degree from the University of Florida in 2005. Her

graduate work has focused on neuropsychology in epilepsy, traumatic brain injury, and

dementia. She is currently attending the Charleston Consortium Internship Program in the

Neuropsychology Track.









the overlapping pairs. A correct choice was reinforced by a "smiley face." Sample instructions

are provided. For the non-overlapping pairs, participants were instructed:

You are going to see pairs of objects on the screen. Press the red key to choose the left
object and the yellow key to choose the right object. Your job is to learn which object is
the winner in each pair. When you pick the winner, a smiley face will appear. You won't
see the smiley face if you pick the incorrect object. Initially you will have to guess which
object is the winner in each pair. Once you find out which object is the winner in a pair,
remember the answer for the next time you see that particular pair.

Prior to training for the overlapping pairs, participants were instructed:

In this next part, you will see new objects. Again your job is to pick the correct object, the
one that produces the smiley face. In this part, a particular object will not be paired with
the same partner each time. Whether an object is correct depends on its partner. As before,
each time a specific pair of objects is presented, the same one will always be correct.
However, if an object is paired with a new partner, it may or may not be correct.

Participants were also provided examples before beginning training for each stimulus set.

Training for each stimulus set consisted of 144 trials, whereby each pair was presented 36 times.

The training procedure was separated into three blocks (Table 2-1). The first and second blocks

consisted of 60 trials. The first training block was "frontloaded" to contain twice as many

representations of two of the pairs, while the second training block was "backloaded" to contain

more representations of the other two pairs. For example, participants viewed 20 instances of

AB and BC, and 10 instances of CD and DE in the first training block for the overlapping set.

During the second block, participants viewed 20 instances of CD and DE, and 10 instances of

AB and BC. According to Heckers and colleagues (2004), the initial front-loading of pairs is

necessary for healthy adults to correctly make inferences during the test. The third training block

consisted of 24 trials containing equal numbers of the four stimulus pairs. Throughout the

training trials, the presentation of the pairs and the position of the two stimuli within each pair

were randomized.









Barr, Hamberger, & Helmstaedter, 2007; Stroup et al., 2003). The risk for cognitive decline is

inversely related to the functional adequacy of the tissue to be rejected in the surgical temporal

lobe (Chelune, 1995). For example, post-surgical patients tend to show a greater loss in verbal

memory if they had better baseline verbal memory functioning, which is indicative of more

functionally adequate underlying tissue (Helmstaedter, 2004; Stroup et al., 2003). Therefore,

neuropsychological testing provides valuable information that can help physicians and patients

inform their decisions about likely outcomes of treatment.

Assessment of Memory Functioning in TLE

Impaired declarative memory is a hallmark symptom of TLE. The material-specific

framework guides our understanding of memory in TLE. Material-specific predictions link

memory for verbal material with the language dominant temporal lobe (usually left) and memory

for nonverbal material with the language non-dominant temporal lobe (usually right; Milner,

1975). Factor analytic results suggest that the constructs of verbal and nonverbal memory are

robust within the TLE population (Davis, Andresen, Witgert, & Breier, 2006).

The literature consistently shows a strong link between impaired verbal memory and left

TLE (Helmstaedter, 2004; Hermann, Seidenberg, Schoenfeld, & Davies, 1997; Strauss et al.,

1995) and left hippocampal atrophy (Griffith et al., 2003; Martin et al., 1999; Trennery, 1996).

In contrast, findings are mixed regarding the association between nonverbal memory and right

TLE. Research generally has not demonstrated a reliable relationship between performance on

nonverbal/visual memory measures and right TLE or right hippocampal volume (Griffith et al.,

2003; Helmstaedter, 2004; Martin et al., 1999; O'Brien, Bowden, Bardenhagen, & Cook, 2003;

Trennery, 1996).

While verbal memory is typically assessed by story-recall or list-learning paradigms, there

is little agreement on which tests best identify nonverbal memory impairments. Currently most









Potential participants were mailed up to three times in order to optimize participation. The

overall response rate to the mailing was 56%. Seven patients were unable to be contacted due to

change in address. Participation in the study was completed by 39% of patients. Twelve patients

responded to the mailing but did not participate for various reasons [distance to travel (n=3),

failed to meet inclusion criteria after additional screening (n=4), declined (n=3), unable to

schedule (n=2)]. Healthy controls were recruited by asking epilepsy patients whether their

family members or friends might be interested in participating and also by community flyer. All

potential participants underwent a 10-minute screening by phone regarding demographic and

medical information relevant to study eligibility.

Assessment Procedures

Eligible candidates were scheduled for testing and assigned a subject number for

identification. Participants were tested at University of Florida, in their home, or at the local

Epilepsy Foundation in their town of residence. The assessment was administered by a graduate

student in psychology or a trained undergraduate research assistant. The duration of the testing

session was three to four hours. Informed consent procedures were completed. Then, a

neuropsychological battery consisting of experimental and traditional measures was

administered. Participants were compensated $10 per hour for their time, and those tested at the

University of Florida were also offered a $3 parking voucher.

Demographic information, including age, gender, ethnicity, handedness, and years of

education, was collected. The epilepsy patients were also asked to provide medical information

and to consent to release information from their treatment in the Comprehensive Epilepsy

Program. Relevant medical information was then collected from the patient's medical record.

Medical variables included age at onset of epilepsy, family history of epilepsy, current

medications, neuroimaging results, seizure laterality and localization, current and previous









that were generated. The initial power analysis was based on the one published study of TI in a

clinical population. This study compared individuals with schizophrenia to nonpsychiatric

controls using nearly identical stimuli and a comparable paradigm (Titone et al., 2004). The TI

effect (d = .80) was large (Cohen, 1988). A power analysis (power = .80) revealed that 63

participants would be needed to detect a large effect in an omnibus three-group test using an

alpha level of .05. Therefore, an optimal sample size would have included at least 20

participants per group. While the control and right ATL groups were close to this target, greater

difficulty was encountered in recruiting patients who had undergone left ATL.

One methodological concern is the verbalizability of stimuli in the task. Nonverbal tests

may be contaminated by verbal encoding (Barr et al., 1997). Studies that support this idea have

shown an association between nonverbal memory and left hippocampal volume in TLE (Griffith

et al., 2003; McConley et al., 2008). In another study, patients with right TLE showed memory

impairment for designs only when their complexity exceeded verbal learning capacity

(Helmstaedter et al., 1995). In the current study, all participants were questioned about the

strategies they used to complete the TI task. Some participants reported applying verbal labels to

the stimulus patterns (e.g., "plaid"), while others reported focusing on the visual patterns and

denied using any verbal strategies. In future studies, it will be essential to generate stimuli that

are difficult to verbally label in order to develop a more pure measure of nonverbal relational

memory.

In addition to considering new stimuli, the TI task could potentially benefit from other

revisions as well. First, it could be beneficial to program the task to ensure that a specific

learning criterion is met during training (e.g., 80%). The advantage to this procedure would be

ensuring that all participants achieve a minimal level of learning, which is important when









Functional imaging studies have indicated that encoding and recall of associative

memories are functions of an integrated hippocampal system (Davachi & Wagner, 2002; Jackson

& Schacter, 2004; Meltzer & Constable, 2005). The behavioral data from the current study also

provide information about the processes involved in relational memory. Groups achieved fairly

comparable levels of learning during training for the overlapping pairs, suggesting that the pairs

were encoded to a similar degree across groups. On the test, patients with right ATL showed a

trend towards poorer performance on the condition that tested memory for overlapping pairs and

showed significantly poorer performance than controls on the condition requiring TI across the

overlapping pairs. This pattern of performance suggests that consolidation and retrieval of

relational information was difficult for patients with right ATL. In particular, the ability to

flexibly use retrieved information and apply it to solve a novel problem was impaired.

It was hypothesized that laterality effects would be observed in the current study, such that

TI for nonverbal information would be more difficult for patients with right ATL than patients

with left ATL. While the task did not clearly discriminate side of surgery, patients with right

ATL did perform more poorly than controls. Past research on TI has not been designed to

address questions of laterality, and hemispheric lateralization observed during functional

imaging is not easily interpreted. Heckers and colleagues (2004) demonstrated right anterior

hippocampal activation during TI with the paradigm adapted for use in the current study.

However, in a later study using a similar paradigm (Zalesak & Heckers, 2009), lateralization

effects were less clear and varied based on the aspect of TI that was studied. Activation was

greater in the right hippocampus for pairs that were more adjacent in the hierarchy, but greater in

the left hippocampus for comparisons of pairs that did not contain end items versus those that









LIST OF REFERENCES


Abrahams, S., Morris, R. G., Polkey, C. E., Jarosz, J. M., Cox, T. C., Graves, M., et al. (1999).
Hippocampal involvement in spatial and working memory: A structural MRI analysis of
patients with unilateral mesial temporal lobe sclerosis. Brain and Cognition, 41, 39-65.

Aggleton, J. P., & Brown, M. W. (1999). Episodic memory, amnesia, and the hippocampal-
anterior thalamic axis. Behavioral and Brain Sciences, 22, 425-444.

Army Individual Test Battery. (1944). Manual of Directions and Scoring. Washington, DC: War
Department, Adjutant General's Office.

Barr, W. B., Chelune, G. J., Hermann, B. P., Loring, D. W., Perrine, K., Strauss, E., et al. (1997).
The use of figural reproduction tests as measures of nonverbal memory in epilepsy surgery
candidates. Journal of the International Neuropsychological Society, 3, 435-443.

Barr, W., Morrison, C., Zaroff, C., & Devinsky, O. (2004). Use of the Brief Visuospatial
Memory Test-Revised (BVMT-R) in neuropsychological evaluation of epilepsy surgery
candidates. Epilepsy & Behavior, 5, 175-179.

Baxendale, S. A., Thompson, P. J., & Paesschen, W. V. (1998). A test of spatial memory and its
clinical utility in the pre-surgical investigation of temporal lobe epilepsy patients.
Neuropsychologia, 36, 591-602.

Benedict, R. H. (1997). Brief Visuospatial Memory Test-Revised. Odessa, FL: Psychological
Assessment Resources.

Brandt, J. & Benedict, R. H. B. (2001). Hopkins Verbal Learning Test-Revised. Odessa, FL:
Psychological Assessment Resources.

Breier, J. I., Plenger, P. M., Castillo, R., Fuchs, K., Wheless, J. W., Thomas, A. B., et al. (1996).
Effects of temporal lobe epilepsy on spatial and figural aspects of memory for a complex
geometric figure. Journal of the International Neuropsychological Society, 2, 535-540.

Brown, M. W., & Aggleton, J. P. (2001). Recognition memory: What are the roles of the
perirhinal cortex and hippocampus? Nature Reviews Neuroscience, 2, 51-61.

Chelune, G. J. (1995). Hippocampal adequacy versus functional reserve: Predicting memory
functions following temporal lobectomy. Archives of Clinical Neuropsychology, 10, 413-
432.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Lawrence Erlbaum Associates.

Cohen, N. J., Poldrack, R. A., & Eichenbaum, H. (1997). Memory for items and memory for
relations in the procedural/declarative memory framework. Memory, 5, 131-178.









with considerably older ages than were represented in the current study (Ryan, Moses, & Villate,

2009). Other demographic and clinical variables, including level of education, age at onset of

epilepsy, duration of epilepsy, and post-surgical seizure freedom, did not show a significant

relationship with accuracy scores.

On reaction times for the test, the only group difference noted was faster performance by

controls than either epilepsy patient group on the Non-overlapping Trained condition. This

finding likely reflects reductions in cognitive and psychomotor speed associated with chronic

temporal lobe epilepsy (Hermann et al., 2006) and with the use of antiepileptic drugs (Motamedi

& Meador, 2004). Moreover, results showed that the non-overlapping conditions may have been

easier for participants, as evidenced by higher accuracy scores (Non-overlapping Trained >

Overlapping Trained, Overlapping Inference) and faster reaction times (Non-overlapping

Trained, Non-overlapping Inference > Overlapping Trained). Since memory demands were less

for these conditions, they may have had a greater likelihood of detecting differences in basic,

speeded processing capacities.

After the computerized task, hierarchical awareness was measured by asking participants

to order the stimuli in the overlapping set. Their ability to demonstrate knowledge of the

hierarchy accounted for 32% of the variance in performance on the TI condition, after

controlling for the effects of age and initial learning. Significant group differences in self-

reported hierarchical awareness were not observed, although twice as many controls (62%) as

right ATL patients (31%) endorsed hierarchical awareness. Libben and Titone (2008) reported

that participants who are aware of the hierarchy may be more likely to use a logic-based strategy

to solve the transitive problem, while participants who are unaware may be more likely to use a

stimulus-driven strategy.









found to perform equivalently to controls on indirect tests of familiar information, but worse on

indirect tests for novel item and associative information (Gooding, Mayes, & van Eijk, 2000).

Relational Memory: Application to Temporal Lobe Epilepsy

The previous section showed how relational memory is characterized by the formation of a

flexible, integrated representation that relies on hippocampal functioning. Strengths of the

relational memory concept include a well-grounded physiological basis and converging

empirical support from studies of animals, healthy adults, and clinical populations. To date,

there have been no studies of relational memory in temporal lobe epilepsy (TLE) using the

experimental paradigms described in the preceding sections.

Several studies are suggestive of relational memory deficits in TLE, although none have

directly contrasted relational memory with memory for items. For instance, patients with left

TLE did not profit in learning a word list when words belonged to schemas about events, while

those with right TLE benefited from these loose associations (Helmstaedter, Gleissner, Di Perna,

& Elger, 1997). In a post-surgical sample, patients with left TLE performed worse than those

with right TLE on memory for word pairs in declarative and nondeclarative formats (Savage,

Saling, Davis, & Berkovic, 2002). Similarly, the extent of surgical resection has been associated

with a decreased ability to learn associations between objects and faces (Weniger, Boucsein, &

Irle, 2004).

Evaluating relational memory in TLE has the potential to provide significant information

about the role of the hippocampus. Through EEG monitoring and neuroimaging techniques, the

complex partial seizures of TLE have been demonstrated to be predominantly of hippocampal

origin (Engel, 1996). The most common structural abnormality is unilateral mesial temporal

sclerosis (Engel, 1996), which is the only neuropathologic finding in the majority of patients

(Trennery, Westerveld, & Meador, 1995). Thus TLE provides a naturally occurring model of









One reason for the poor performance of conventional nonverbal memory tests in this

context is that they focus primarily on object/item memory, a function that may not specifically

reflect the hippocampal contribution to memory. Other possible reasons exist (see below).

Nonverbal tests that tap aspects of spatial memory may be more promising. For instance, several

researchers have proposed alternative scoring methods for the Rey Complex Figure. Loring,

Lee, and Meador (1988) showed that qualitative scoring criteria assessing distortion and

misplacement errors discriminated between right and left TLE, differences that were not evident

when traditional scoring criteria were used. These results were replicated by Frank and

Landeira-Fernandez (2008). Likewise, Breier and colleagues (1996) found that a spatial scoring

system for the Rey Complex Figure was more sensitive to right hippocampal dysfunction than a

figural scoring system; however, subsequent studies have not detected differential sensitivity to

right TLE using this scoring (Kneebone, Lee, Wade, & Loring, 2007; McConley et al., 2008).

Experimental spatial memory tests are also promising. Patients with right TLE have shown

worse performance than those with left TLE on a task requiring them to remember spatial details

in complex scenes (Baxendale, Thompson, & Paesschen, 1998) and on a task requiring them to

change position while remembering the location of hidden objects in a spatial array (Abrahams et

al., 1999).

Improving Detection of Memory Dysfunction in Right TLE

As indicated above, there are several possible explanations for the failure to demonstrate a

consistent link between nonverbal memory performance and right TLE. Some researchers

suggest that nonverbal memory may not be as strictly lateralized or localized as verbal abilities,

which could help to account for the findings (Helmstaedter & Kurthen, 2001; Loring et al.,

2007). Other explanations focus on test characteristics that may contribute to their insensitivity

to nonverbal memory dysfunction.









Table 3-2. Transitive inference accuracy scores as percent correct by group
Right ATL Left ATL Control
(n=16) (n=8) (n=24)
Training: Non-overlapping pairs 77.9 (20.2) 81.4 (10.5) 88.1 (15.6)
Block 1 66.0 (16.6) 68.7 (12.7) 78.9 (16.3)
Block 2 82.2 (23.5) 83.9 (14.3) 91.2 (18.4)
Block 3 85.4(24.4) 91.7(8.6) 94.1 (15.0)
Training: Overlapping pairs 74.6 (15.6) 73.0 (10.5) 81.7 (12.9)
Block 1 69.5(13.1) 69.8(8.0) 73.0(14.9)
Block 2 76.3 (18.4) 71.1 (12.3) 85.8 (15.2)
Block 3 77.9 (18.2) 78.1 (17.2) 86.4 (13.3)
Test
Non-overlapping learned 85.0 (22.5) 90.6 (12.6) 94.1 (15.7)
Non-overlapping inference 82.0 (28.4) 82.8 (23.6) 92.8 (20.6)
Overlapping learned 76.2 (16.9) 82.2 (17.5) 87.2 (15.3)
Overlapping inference 71.2 (22.1) 82.2 (23.7) 89.1 (15.5)
Note. Data presented as mean (standard deviation) of percentage correct, with 50% representing
chance performance and 100% representing error-less performance.









rely on stimulus novelty or familiarity (Preston & Gabrieli, 2002). Medial temporal structures

implicated in more recognition-based "item" memory include the perirhinal cortex (Aggleton &

Brown, 1999; Brown & Aggleton, 2001; Koehler, Danckert, Gati, & Menon, 2005) and

entorhinal and parahippocampal cortices (Davachi & Wagner, 2002). The distinction between

item and relational memory is subtle but important, and focuses on whether the stimulus has

been encoded as an item, which is a unitized whole, or as a relational representation, which

preserves the constituent elements as well as the larger structure. It is important to recognize that

most clinical memory tests are not designed to be sensitive to this distinction.

Paradigms to Assess Relational Memory

Several paradigms have been developed to specifically assess relational memory. Two of

the most widely used paradigms, paired associate learning and transitive inference (TI), both

incorporate rapid, incidental learning.

Paired associate learning

One paradigm to assess relational memory involves learning paired associates (e.g.,

Giovanello, Schnyer, & Verfaellie, 2004; Meltzer & Constable, 2005). The subject is trained on

a series of pairs, A-B, C-D, E-F, where A-F can be words or pictures. During a test phase, the

subject is presented with intact pairs (e.g., A-B), novel pairs (e.g., Y-Z), and recombined pairs

(e.g., A-F; items presented previously, but not together). Due to their similar exposure histories,

items in intact and recombined pairs are equally familiar to subjects and so differ only in whether

they were previously associated. Performance differences on intact versus recombined pairs

provides an index of relational memory, while performance differences on recombined versus

novel pairs provides a measure of item memory. Thus, the paired associate paradigm allows

unique assessment of relational and item memory and eliminates the confound of familiarity on

performance. This paradigm has demonstrated sensitivity to hippocampal function in amnesic








Non-overlapping pairs


a b c d e f g h





Overlapping pairs


A B B C C D D E


i X1 >


Figure 2-1. Transitive inference stimuli.









did. The lateralization of TI is deserving of further study using functional imaging and

behavioral paradigms.

More broadly, TI reflects relational memory processing because it examines whether

representations that share common elements have been linked in memory to solve new problems

(Eichenbaum, 2000). These results provide support for the relational memory account of

hippocampal function in a new population, patients with complex partial epilepsy of temporal

lobe origin. The findings also suggest a laterality effect for TI, such that relational memory

processing for visual/nonverbal information is difficult for TLE patients with right hippocampal

pathology and side of surgery.

Clinical Applications of Transitive Inference

A secondary aim of the study was to compare the clinical utility of the TI task to

conventional neuropsychological measures. The TI score (i.e., accuracy on inferences for

overlapping pairs) was chosen to represent performance because it appeared most sensitive to

differential performance in TLE. Material specific-predictions link memory for verbal material

to the language dominant temporal lobe (usually left) and memory for nonverbal material to the

language non-dominant temporal lobe (usually right; Milner, 1975). Accordingly, the TI task

was compared to the standard measure of nonverbal memory, the Rey Complex Figure Test.

In the current study, results support the strong link between impaired verbal memory and

left TLE, which has been consistently demonstrated in the literature (Helmstaedter, 2004;

Hermann et al., 1997, Strauss et al., 1995). Scores from the HVLT-R and the WMS-III Verbal

Paired Associates discriminated patients with left ATL from those with right ATL. Both

measures assess verbal list learning capacity and showed large effect sizes in discriminating

between groups. The largest effect size (r2 = .30) was observed for the WMS-III Verbal Paired

Associates Total score, which suggests that memory for word pairs as opposed to single words









CHAPTER 2
METHODS

Participants

Participants included patients with intractable temporal lobe epilepsy (TLE; n=27), who

had undergone a standard anterior temporal lobectomy (ATL), and healthy controls (n=24).

Participants were 18 years of age or older. Exclusion criteria were: 1) history of developmental

disability or mental retardation resulting in Wechsler Abbreviated Scale of Intelligence (WASI;

Psychological Corporation, 1999) Full-Scale IQ < 70; 2) history of Axis I psychiatric disturbance

resulting in hospitalization; 3) cranial radiation or chemotherapy treatment (within 1 year); or 4)

other neurological illness (e.g., cerebrovascular disease, brain tumor, severe traumatic brain

injury). Controls were required to be free of any neurological disease. After completion of

testing, three epilepsy patients were excluded from further analyses. Two were eliminated due to

concerns about comprehension since English was a second language, and another was excluded

due to current psychiatric symptoms that resulted in failure to complete the testing session.

Procedures

Recruitment Strategy

The study was approved by the University of Florida Institutional Review Board (#240-

2007). Epilepsy patients were recruited from the Comprehensive Epilepsy Program at the

University of Florida. Patients who had undergone ATL from January 2000 to January 2008

were pre-screened in medical records to determine if they met inclusion/ exclusion criteria for

the study. Seventy-seven patients (38 left ATL, 39 right ATL) met eligibility criteria and were

contacted by mail. A letter was sent that informed them of the opportunity to participate in a

research study and briefly described the study. Enclosed was an addressed, stamped postcard

that could be returned, should they desire additional information or want to volunteer.









Kroll, N. E., Knight, R. T., Metcalfe, J., Wolf, E. S., & Tulving, E. (1996). Cohesion failure as a
source of memory illusions. Journal of Memory and Language, 35, 176-196.

Lee, T. M., Yip, J. T., & Jones-Gotman, M. (2002). Memory deficits after resection from left or
right anterior temporal lobe in humans: A meta-analytic review. Epilepsia, 43, 83-91.

Libben, M., & Titone, D. (2008). The role of awareness and working memory in human
transitive inference. Behavioural Processes, 77, 43-54.

Loring, D. W. (1997). Neuropsychological evaluation in epilepsy surgery. Epilepsia, 38(Suppl.
4), S18-23.

Loring, D. W., Barr, W., Hamberger, M., & Helmstaedter, C. (2007). Neuropsychology
evaluation adults. In T. A. Pedley & J. Engel, Jr. (Eds.), Epilepsy: A Comprehensive
Textbook (2nd ed.) (pp. 1057-1066). Philadelphia: Lippincott, Williams, & Wilkins.

Loring, D. W., Lee, G. P., & Meador, K. J. (1988). Revising the Rey-Osterrieth: Rating right
hemisphere recall. Archives of Clinical Neuropsychology, 3, 239-247.

Martin, R. C., Hugg, J. W., Roth, D. L., Bilir, E., Gilliam, F. G., Faught, E., et al. (1999). MRI
extrahippocampal volumes and visual memory: Correlations independent of MRI
hippocampal volumes in temporal lobe epilepsy patients. Journal of the International
Neuropsychological Society, 5, 540-548.

Mayes, A. R., Holdstock, J. S., Isaac, C. L., Montaldi, D., Grigor, J., Gummer, A., et al. (2004).
Associative recognition in a patient with selective hippocampal lesions. Hippocampus, 14,
763-784.

McConley, R., Martin, R., Palmer, C. A., Kuzniecky, R., Knowlton, R., & Faught, E. (2008).
Rey Osterrieth complex figure test spatial and figural scoring: relations to seizure focus
and hippocampal pathology in patients with temporal lobe epilepsy. Epilepsy and
Behavior, 13, 174-177.

McGonigle, B., & Chalmers, M. (1977). Are monkeys logical? Nature, 267, 694-696.

Meltzer, J. A., & Constable, R. T. (2005). Activation of human hippocampal formation reflects
success in both encoding and cued recall of paired associates. Neuroimage, 24, 384-397.

Meyers, J. E., & Meyers, K. R. (1995). Rey Complex Figure Test and Recognition Trial manual.
Odessa, FL: Psychological Assessment Resources.

Milner, B. (1975). Psychological aspects of focal epilepsy and its neurosurgical management.
Advances in Neurology, 8, 299-321.

Motamedi, G. K., & Meador, K. J. (2004). Antiepileptic drugs and memory. Epilepsy and
Behavior, 5, 435-439.









Therefore, ANCOVAs were run for these three test conditions, with group as the between-

subjects factor, accuracy score as the dependent variable, and age as the covariate. Age was a

significant covariate in all the analyses (p < .05). Only on the Overlapping Inference condition

was a main effect of patient group observed, F(2, 44) = 4.54, p = .02, r2 = .20. Again,

Bonferroni adjusted post-hoc tests showed that patients with right ATL performed worse than

controls, t(44) = 3.00, p = .01, on this TI condition.

In addition, seizure freedom was considered as a potential influence on test performance

(Table 3-3). Two left ATL patients and two right ATL patients continued to have seizures after

surgery, which is a less successful outcome and indicates that epileptogenic tissue remains in

their brains and is potentially disturbing cognitive functions. Test performance was re-examined

when these four patients were excluded from the sample. ANCOVAs with group as the

between-subjects factor, accuracy score for the test as the dependent variable, and age as the

covariate revealed the same pattern of results observed in the larger sample. These analyses,

coupled with the non-significant correlations between seizure freedom and test performance,

suggest that including these patients in the sample should not have a major impact on the results.

In light of the poor performance of right ATL patients on the Overlapping Inference

condition, an examination of individual inference pairs was made (Figure 3-4). The critical

inference pair, BD, contained stimuli that had both been reinforced 50% of the time during

training. The remaining four inference pairs contained end items and would be expected to have

a lesser degree of difficulty. Thus, the largest disparity between groups would be expected on

the BD pair. In view of the small sample and skewed distributions of some scores, performance

on individual pairs was examined with the Kruskal-Wallis H test, which is the nonparametric

analogue to the between-subjects ANOVA. Group was the independent variable, and accuracy









ACKNOWLEDGMENTS

I thank my dissertation chair, Dr. Russell Bauer, for being a great mentor during my

graduate training. I am very grateful for the exceptional guidance and support that he provided

through all aspects of this project, from developing the concept to data collection and analysis. I

would also like to express my sincere gratitude to the other members of my dissertation

committee, Michael Marsiske, David Loring, Eileen Fennell, and Steven Roper. I especially

thank Dr. Michael Marsiske for his invaluable help with statistical issues and his insight into

working with my experimental data. Lastly, I am grateful for the love and support of my

wonderful husband and family throughout my graduate education.









which resulted in low sensitivity. For instance, at the 75% cut score on TI, positive predictive

power was 89%, but only half of patients with right ATL were detected. Higher cut-offs yielded

unacceptably high rates of false positives.

Taken together, these findings provide limited support for our secondary hypothesis that

the TI task would be more sensitive to right ATL than the conventional nonverbal memory

measure. The TI task detected differences between right ATL patients and controls, while the

Rey Complex Figure Test did not. However, neither measure discriminated between epilepsy

groups with sufficient accuracy to be used with confidence at the single-patient level.

Previous research has not compared TI to results of standard neuropsychological testing.

Partial correlations controlling for age provided some evidence of convergent and discriminant

validity for the experimental task. All significant correlations were in the small to moderate

range (r = .3 to .4). TI showed a moderate correlation with an estimate of intellectual

functioning, suggesting that it taps some general cognitive abilities. TI also showed modest

correlations with the Rey Complex Figure Test recall measures, which provides some degree of

convergent validity. Both transitive and non-transitive inference scores were associated with a

measure of contextual, verbal recall and Trail Making Test Part B. The relationships between TI

and other measures of verbal memory, language, attention, and executive function were not

significant, which provides evidence of discriminant validity. Non-transitive inference, however,

showed low correlations with several measures of attention and executive function. This

relationship was unexpected and may suggest a frontal lobe contribution.

Limitations and Future Directions

A limitation of the current study is the relatively small sample size that may have resulted

in some of the analyses being underpowered. In light of the limited power, several exploratory

analyses were conducted that were, strictly speaking, not statistically justified using the models









On the test, patients who had undergone right ATL performed significantly worse than

healthy controls on TI for visual information. Left ATL patients performed in the intermediate

range; however, the task did not discriminate between patients based on side of surgery. Results

provide some evidence of a laterality effect and suggest that TI may be sensitive to

hippocampally-mediated memory function. There is a clear need for better neuropsychological

measures to assess language non-dominant (usually right) temporal lobe function, given the poor

sensitivity and specificity of current tests. In this study, the conventional nonverbal memory

measure, the Rey Complex Figure Test, did not discriminate between groups. The TI task and

the conventional nonverbal measure yielded similar operating characteristics with good positive

predictive power but poor sensitivity. The TI task showed modest clinical promise, and

modifications that may improve its clinical utility are suggested.









The nature of the test stimuli, in particular their level of abstractness, should be considered.

Many authors have raised concerns about the verbalizability of stimuli, such that nonverbal tests

may be contaminated by verbal encoding (Barr et al., 1997). Griffith and colleagues (2003)

found that the left hippocampus explained a significant portion of the variance in nonverbal

memory performance, which may be considered to support this contention. Similarly, McConley

and colleagues (2008) noted modest correlations between left hippocampal volume and figural

reproduction performance. Moreover, Helmstaedter, Pohl, and Elger (1995) found that patients

with right TLE showed memory impairment for designs only when their complexity exceeded

verbal learning capacity. An additional concern is that many measures involve reproduction of

designs, which is confounded by motor and constructional abilities. A recognition format would

bypass this concern.

Another explanation for the mixed findings relates to the type of nonverbal memory

assessed. Results described previously indicate that spatial memory tests may be more closely

tied to language non-dominant hippocampal function than tests of figural or object memory.

These findings could be interpreted in the context of relational memory theory. As spatial

memory involves integrating relationships to represent the environment, it could be considered a

subset of relational memory (Cohen et al., 1999). The success of some spatial measures in the

identification of neuropsychological morbidity in TLE suggests that other relational memory

paradigms may also have clinical assessment value.

Most studies on TLE are conducted in clinical settings and use commercially available

assessments rather than experimental paradigms. However, commercial tests may not provide

the most sensitive measure of hippocampally-mediated memory function. Measures, such as TI,



































To my family









LIST OF TABLES


Table page

2-1 Protocol for the transitive inference task ..................................... ........... ........ ....... 34

3-1 Demographic and clinical characteristics by group .......................................................51

3-2 Transitive inference accuracy scores as percent correct by group................................. 52

3-3 Correlations between TI test, demographic characteristics, and clinical variables ..........55

3-4 Transitive inference reaction times by group.............................. ...............55

3-5 Neuropsychological test performance by group ..................................... .................58

3-6 Partial correlations between transitive and non-transitive inference and performance
on standard neuropsychological tests after controlling for age ......................................59









129) = 2.86, p = .01, r12 = .12. Decomposing this interaction revealed that control participants

had faster reaction times on the Non-overlapping Trained condition compared to either patients

with right ATL (p = .04) or left ATL (p = .04). Participants exhibited comparable reaction times

on conditions involving the overlapping stimulus set.

In addition, reaction times on the test were analyzed to ascertain if the TI effect reported by

Heckers and colleagues (2004) could be replicated. The TI effect would be demonstrated by an

interaction between sequence and inference, such that reaction times would be slowest on

responses requiring inference across the overlapping pairs. Reaction times on the test were

analyzed using 2 (Sequence Type: overlapping, non-overlapping) x 2 (Inference Type: present,

absent) repeated measures ANOVAs. In the control group, a main effect of sequence was

observed, F(1, 22) = 28.61, p < .001, r12 = .56, such that controls showed significantly faster

reaction times on the non-overlapping compared to the overlapping pairs. Neither the main

effect of inference [F(1, 22) = 1.01, p = .33, r2 = .04] nor the interaction between sequence and

inference was significant [F(1, 22) =.59, p = .45, 12 = .03]. This analysis was also conducted in

the epilepsy sample. In the epilepsy patients, a main effect of sequence was observed, F(1, 22) =

6.13, p < .02, r2 = .22, such that patients also showed significantly faster reaction times on the

non-overlapping compared to the overlapping pairs. Again neither the main effect of inference

[F(1, 22) = .16, p = .69, r2 = .01] nor the interaction between sequence and inference was

significant [F(1, 22) =.13,p = .72, r2 = .01].

Aim 2: Comparison of Transitive Inference and Standard Neuropsychological Tests

The secondary aim of the study was to compare the clinical utility of the TI task to

established clinical memory tests. It was hypothesized that the TI task would provide more

sensitive detection of language non-dominant side of surgery than standardized

neuropsychological tests of memory. Results from the previous section showed that accuracy on









Standard Neuropsychological Tests

Conventional neuropsychological measures were used to assess general intellectual

functioning, verbal and nonverbal memory, language, attention, and executive function. The

following tests are peer-reviewed, well-validated measures to assess these domains. All tests

were administered and scored according to standardized procedures outlined in the test manuals.

Resulting performances were demographically corrected to remove inherent differences due to

age. In addition, the normative data provided corrections for different educational levels for the

measures indicated by an asterix.

1) Wechsler Abbreviated Scale of Intelligence (WASI; Psychological Corporation, 1990).

The Vocabulary subtest and the Matrix Reasoning subtest were administered. In Vocabulary, the

examinee is asked to orally define words. In Matrix Reasoning, the examinee is asked to select a

pattern to complete an abstract design. Performance on these two subtests can be combined to

yield an estimate of general intellectual functioning.

2) Hopkins Verbal Learning Test-Revised (HVLT-R; Brandt & Benedict, 2001). The

HVLT-R is a measure of the processes involved in learning and remembering verbal

information. The examinee's ability to learn a 12-word list over three trials is examined. Free

recall of the list is assessed after a 20- to 25-minute delay, followed by a recognition trial.

3) Wechsler Memory Scale-Third Edition selected verbal subtests (WMS-III; Wechsler,

1997b). In Logical Memory, participants are read two brief stories and asked to retell them

immediately and after a 30-minute delay. In Verbal Paired Associates, participants are asked to

learn a list of eight abstract word pairs over four learning trials. During the immediate and 30-

minute delayed recall, the examiner reads the first word in each pair, and the examinee is asked

to recall the second word. Both subtests also include a recognition trial.









The current study utilized a TI paradigm closely modeled after the task developed by

Heckers and colleagues (2004). The results partially replicated their behavioral data. In their

healthy, young adult sample, participants achieved a mean of greater than 90% on all test

conditions. Healthy controls in the current study achieved similar mean scores, ranging from 87

to 94% on the various test conditions, while the scores of epilepsy patient tended to be lower.

Regarding reaction time data, Heckers and colleagues (2004) noted a main effect of sequence

(i.e., reaction times faster for non-overlapping than for overlapping pairs), a main effect of

inference (i.e., reaction times quicker for responses that did not require an inference), and an

interaction between sequence and inference (i.e., the TI effect: reaction times slowest on

responses requiring an inference across overlapping pairs). In the current study, the effect of

sequence was replicated, but no main effect of inference nor interaction effect was observed in

either control participants or epilepsy patients.

Transitive Inference and Cognitive Neuroscience Research

The post-surgical ATL population provides a model to study hippocampal contributions to

memory, and the current results suggest a link between hippocampal function and TI. Functional

imaging studies have identified a distributed neural network involved in TI judgments, including

areas of the cortex, hippocampus, and thalamus (Heckers et al., 2004; Zalesak & Heckers, 2009).

This research has consistently highlighted the integral role of the hippocampus in TI and has

demonstrated selective hippocampal activation during TI of geometric designs (Heckers et al.,

2004; Zalesak & Heckers, 2009), faces (Nagode & Pardo, 2002), and face-house pairings

(Preston et al., 2004). The translational paradigm of TI is also supported in studies of animals,

which showed that hippocampally-lesioned rodents exhibit impaired TI (Devito et al., in press;

Dusek & Eichenbaum, 1997).









While the majority of studies have focused on encoding processes, several have examined

retrieval processes. Encoding and recall of associative memories are functions of an integrated

hippocampal system (Meltzer & Constable, 2005). Research has revealed hippocampal

activation during recognition of relational information, but not item information (Preston et al.,

2004; Yonelinas, Hopfinger, Buonocore, Kroll, & Baynes, 2001). Furthermore, some studies

support the idea that relational encoding is associated with anterior hippocampal activity, while

retrieval is associated with posterior hippocampal activity (Meltzer & Constable, 2005; Prince,

Daselaar, and Cabeza; 2005). In both studies, the overlap of encoding and retrieval effects was

maximal in the middle of the longitudinal extent of the hippocampus, near the CA3 area.

Neuroimaging studies also provide evidence of selective hippocampal activation, relative

to other temporal regions, during TI tasks (Nagode & Pardo, 2002; Preston et al., 2004) and other

relational memory tasks (Koehler et al., 2005). For instance, Heckers and colleagues (2004)

identified a distributed network of brain regions, including the pre-supplementary motor area,

bilateral frontal cortex, bilateral parietal cortex, bilateral posterior temporal cortex, and pulvinar,

involved in TI for overlapping visual (nonverbal) stimulus pairs. Importantly, difference

analysis demonstrated selective right anterior hippocampal activation during TI and bilateral

activation in the anterior parahippocampal gyrus during other task conditions. Moreover, in a

similar study (Zalesak & Heckers, 2009), greater hippocampal activation was associated with

more cognitively demanding aspects of TI. Specifically, greater right hippocampal activation

was observed for inferences about pairs derived from more adjacent items in the hierarchy, while

greater left hippocampal activation was associated with inference pairs that did not contain end

items versus those that did.









Table 3-3. Correlations between TI test, demographic characteristics, and clinical variables
Non- Non-
overlapping overlapping Overlapping Overlapping
trained inference trained inference
Age -.25 -.33* -.47** -.46**
Years of education .18 .25 .17 .20
Age at onset (yrs.) -.30 -.38 -.10 -.13
Duration of epilepsy (yrs.) -.06 -.13 -.26 -.18
Seizure freedom .26 .34 .02 .22
Note. Square-root transformed accuracy scores were used in the Pearson bivariate correlations.
*p<.05. **p<.01.

Table 3-4. Transitive inference reaction times by group


Training: Non-overlapping pairs

Block 1

Block 2

Block 3

Training: Overlapping pairs

Block 1

Block 2

Block 3

Test

Non-overlapping learned

Non-overlapping inference

Overlapping learned

Overlapping inference


Right ATL
(n=16)
1177.8 (473.9)

1470.5 (587.7)

1143.2 (474.2)

1075.1 (445.4)

1436.8 (476.5)

1644.8 (646.0)

1411.4(512.6)

1366.7 (488.9)



1163.5 (336.6)

1217.0 (463.7)

1343.4 (510.7)

1404.3 (577.8)


Left ATL
(n=8)
1339.9 (294.9)

1592.4 (321.2)

1272.9 (257.9)

1205.7 (363.7)

1377.6 (286.5)

1706.7 (349.6)

1310.6(302.3)

1223.7 (343.4)



1243.4 (355.7)

1262.4 (575.6)

1261.4 (363.7)

1148.9 (341.2)


Control
(n=24)
908.7 (295.7)

1260.1 (506.6)

905.4 (360.9)

788.5 (235.5)

1147.2 (364.7)

1304.9 (469.0)

1133.0 (369.7)

1058.7 (344.6)



865.4 (323.6)

871.1 (544.6)

1210.7 (524.0)

1156.0 (524.3)


Note. Data presented as mean (standard deviation) of median response time in milliseconds for
correct responses.









ATL would demonstrate significant difficulty making inferences about visual stimulus pairs

from the overlapping set. Latency scores also provided information on relational memory

processing. Based on the findings of Heckers and colleagues (2004), significantly longer

latencies were expected on responses for overlapping vs. non-overlapping pairs and for

responses requiring an inference, with latencies most pronounced on inferences about

overlapping pairs.

Post-test assessment

Following the computer-based task, it was assessed whether participants were explicitly

aware of the hierarchical nature of the stimuli (similar to Titone, Ditman, Holzman, Eichenbaum,

& Levy, 2004). Participants were presented with two hand-held stimulus cards representing B

and D and asked which would be the winner if they were paired. They were then given five

cards representing all the stimuli from the overlapping set and asked to rank order them

according to "dominance" (i.e., which stimulus was most likely to be a winner). The

experimenter provided no further information, allowing participants to decide what attribute to

use to order the cards.

Each stimulus in the arrangement was then assigned a difference score, which was derived

by subtracting its correct position (1, 2, 3, 4, 5) from the position the participant selected (1, 2, 3,

4, 5) and taking the absolute value of the result. For example, a stimulus in the correct position

would be assigned a difference score of zero, whereas a stimulus in the third position that should

be in the first position would be assigned a difference score of two. Difference scores were

summed to yield a total hierarchical awareness score (range 0-12). Lastly, participants were

asked whether they were aware that the stimuli were hierarchically organized and to describe

their strategies for solving the task.



































2009 Marie D. Barker









Van der Jeugd, A., Goddyn, H., Laeremans, A., Arckens, L., D'Hooge, R., & Verguts, T. (2009).
Hippocampal involvement in the acquisition of relational associations, but not in the
expression of a transitive inference task in mice. Behavioral Neuroscience, 123, 109-114.

von Fersen, L., Wynne, C. D., Delius, J. D., & Staddon, J. E. (1991). Transitive inference
formation in pigeons. Journal ofExperimental Psychology: Animal Behavior Processes,
17,334-341.

Wechsler, D. (1997a). Wechsler Adult Intelligence Scale- III manual. San Antonio: The
Psychological Corporation.

Wechsler, D. (1997b). Wechsler Memory Scale-Ill manual. San Antonio: The Psychological
Corporation.

Weniger, G., Boucsein, K., & Irle, E. (2004). Impaired associative memory in temporal lobe
epilepsy subjects after lesions of hippocampus, parahippocampal gyms, and amygdala.
Hippocampus, 14, 785-795.

Yang, J., Weng, X., Guan, L., Kuang, P., Zhang, M., Sun, W., et al. (2003). Involvement of the
medial temporal lobe in priming for new associations. Neuropsychologia, 41, 818-829.

Yonelinas, A. P., Hopfinger, J. B., Buonocore, M. H., Kroll, N. E., & Baynes, K. (2001).
Hippocampal, parahippocampal and occipital-temporal contributions to associative and
item recognition memory: an fMRI study. Brain Ilm.ging. 12, 359-363.

Zalesak, M., & Heckers, S. (2009). The role of the hippocampus in transitive inference.
Psychiatry Research: XNei inuig.iu'. 172, 24-30.









Studies of amnesia

Neuropsychological evidence also supports the relational memory account. Patients with

amnesia due to hippocampal damage have shown selective difficulty with memory for

associations (Kroll et al., 1996) and for configural learning, similar to the childhood "rock,

paper, scissors" game (Rickard & Grafman, 1998). While a few studies have failed to replicate

these results (Stark, Bayley, & Squire, 2002; Stark & Squire, 2003), potentially due to

characteristics of the medial temporal lobe damage in their participants, these findings have been

confirmed in a number of studies.

For instance, impaired between-item associations (e.g., face-face, face-word) with

preserved single item recognition has been observed in six patients with hippocampal lesions

(Turriziani, Fadda, Caltagirone, & Carlesimo, 2004). Likewise, Mayes and colleagues (2004)

reported that a patient with bilateral hippocampal lesions was impaired on memory for cross-

modal associations (e.g., object-location, face-name), but not memory for individual items or

intra-item associations. Hippocampal amnesics experience problems remembering relations

among items, even with very short delays (Hannula, Tranel, & Cohen, 2006; Olson, Page,

Moore, Chatterjee, & Verfaellie, 2006).

The impairment in relational memory is not confined to declarative memory tasks and

extends into the nondeclarative domain as well. Priming, a measure of nondeclarative memory,

is demonstrated when previous exposure to a stimulus leads to a faster reaction time or improved

accuracy when the stimulus is shown again. Investigators have observed impaired priming for

new associations in patients with amnesia. For instance, amnesics with focal medial temporal

lobe lesions demonstrated impaired priming both for associations between two words and for

associations between two features of a single stimulus, but showed preserved item priming

(Yang et al., 2003). In a meta-analysis on implicit memory in organic amnesia, amnesics were









Table 2-1. Protocol for the transitive inference task
Part 1: Training for non-overlapping pairs
Stimuli pairs Eight ovals with geometric, patterned fills
Combined in pairs, with one stimulus in pair arbitrarily designated as winner
(a>b, c>d, e>f, g>h)


Parameters


Blocks


left/right position counterbalanced
random presentation within each block
feedback on every trial such that selecting winner results in smiley face
Block 1: "front-loaded," 60 trials
ab and cd appear 20 times each; ef and gh appear 10 times each
Block 2: "back-loaded," 60 trials
ef and gh appear 20 times each; ab and cd appear 10 times each
Block 3: "balanced," 24 trials
ab, cd, ef, gh appear 6 times each


Part 2: Training for overlapping pairs
Stimuli pairs Five pentagons with geometric, patterned fills
Combined to form a series of overlapping pairs in a hierarchy
(A>B, B>C, C>D, D>E)


Parameters


left/right position counterbalanced
random presentation within each block
feedback on every trial such that selecting winner results in smiley face


Block 1: "front-loaded," 60 trials
AB and BC appear 20 times each; CD
Block 2: "back-loaded," 60 trials
CD and DE appear 20 times each; AB
Block 3: "balanced," 24 trials
AB, BC, CD, DE appear 6 times each


Part 3: Test
Stimuli pairs


Parameters

Blocks


and DE appear 10 times each

and BC appear 10 times each


previously viewed non-overlapping and overlapping pairs
5 novel pairings from non-overlapping set: a>d, a>f, c>f, c>h, e>h
5 novel pairings from overlapping set: A>C, A>D, B>D, B>E, C>E
left/right position counterbalanced
no reinforcement
Block 1 and Block 2:
Each block is 80 trials divided into sets of 10 trials of a particular type
(Non-Overlapping Trained, Non-Overlapping Inference, Overlapping Trained,
Overlapping Inference)


Blocks









seizure frequency, Wada memory support, cerebral speech dominance, date of surgery, type of

surgery, and post-surgical outcome.

Measures

Transitive Inference Task

The transitive inference (TI) task was presented on a laptop computer using stimulus-

presentation software (E-prime). It was closely modeled after the paradigm developed by

Heckers and colleagues (2004). They demonstrated selective right hippocampal activation

during functional imaging of the TI condition, while other task conditions implicated medial

temporal areas, such as the anterior parahippocampal gyms. Their stimuli (Heckers et al., 2004)

were adapted for use in the current study. They consist of thirteen visually distinct pattern fills

created from Corel Draw, which were selected to be of similar levels of visual interest (Figure 2-

1). The patterns were randomly assigned to pairs of ellipses or pentagons.

The stimuli composed a series of four non-overlapping visual stimulus pairs (which will be

represented in text by lower case letters a>b, c>d, e>f, g>h) and a series of four overlapping

visual stimulus pairs (which will be represented in text by uppercase letters A>B, B>C, C>D,

D>E). During training, participants learned the "winner" in each pair. They were then tested on

their ability to recollect the correct response for previously seen pairs and to infer the correct

response for novel pairings. More detailed information on training and testing procedures

follows and is also summarized in Table 2-1.

Training

The training was designed to ensure that participants would learn the correct response for

each pairing and would also be likely to hierarchically encode the overlapping stimulus set

(Heckers et al., 2004). Training was conducted first on the non-overlapping pairs and then on









LIST OF FIGURES


Figure p e

2-1 T ransitive inference stim uli ....................................................................... ..................35

3-1 Accuracy for non-overlapping pairs by training block .............. ...................................53

3-2 Accuracy for overlapping pairs by training block .................................. ...............53

3-3 A accuracy by test condition............................................. ................... ............... 54

3-4 Accuracy for overlapping pairs requiring transitive inference .......................................54

3-5 Reaction time for non-overlapping pairs by training block.............................................56

3-6 Reaction time for overlapping pairs by training block ............................................... 56

3-7 R action tim e by test condition................. ............................ ................. ............... 57

3-8 ROC curves for the prediction of right anterior temporal lobectomy.............................57









TI was selected as the primary measure for the current study because it appears to uniquely

capture the role of the hippocampus in encoding relationships among multiple stimuli, rather

than simply encoding relationships between two simultaneously presented items as in paired

associates. It extends the concept of paired associates to evaluate relationships across pairs. In

addition, TI may be better representative of daily demands on memory processing, as this

paradigm highlights the application of memory in a novel situation. The ability to utilize

information learned across situations, to make inferences, and to apply that knowledge in a novel

situation is critical for higher cognitive function.

Converging Evidence for the Role of the Hippocampus in Relational Memory

The concept of relational memory adds significantly to our understanding of how the

hippocampus is involved in memory. Studies utilizing the paradigms described in the preceding

section have provided strong support for the relational memory account in the animal,

neuroimaging, and neuropsychological literatures.

Animal models

In a review of animal studies, Eichenbaum (2000) reported that deficits in well-validated

tasks used in the animal memory literature, such as a water maze and TI, reflect problems in

relational memory. Several studies have demonstrated relational memory deficits in rodents with

hippocampal lesions. For instance, Fortin, Agster, and Eichenbaum (2002) showed that

hippocampally-lesioned rats were impaired on learning the sequential order of a series of odors,

but not on recognition of previously learned odors. Similarly, Dusek and Eichenbaum (1997)

found that hippocampal disconnections from cortical and subcortical pathways selectively

impaired TI in odor learning. Later studies have attempted to pinpoint the role of the

hippocampus. The hippocampus may help to acquire the underlying representations necessary

for TI in rodents (Van der Jeugd et al., 2009). On a different note, Devito, Kanter, and









Mauchly's test was significant (p < .05), suggesting a violation of the sphericity assumption.

Thus, Greenhouse-Geisser df corrections are reported. A significant effect of condition was

observed, F(1.76, 79.06) = 6.83, p = .003, 12 = .13. Bonferroni adjusted post-hoc tests revealed

greater accuracy on the Non-overlapping Trained condition than on either Overlapping Trained

(p = .01) or Overlapping Novel (p = .02) conditions (Table 3-2). The main effect of group was

also significant, F(2, 45) = 3.94, p = .03, 12 = .15. Bonferroni adjusted post-hoc analyses

showed that patients with right ATL performed more poorly than controls (p = .03) on the task

(Figure 3-3). The interaction between group and condition was not significant, F(3.51, 79.06) =

.73,p = .56, 1= .03.

The primary hypothesis was that patients with right ATL would show selective difficulty

with the TI condition (i.e., inferences about overlapping pairs). Therefore, given the limited

power (.21) of the preceding analysis to detect an interaction, exploratory one-way ANOVAs

were conducted to further evaluate performance on the test conditions. Group (right ATL, left

ATL, control) was the between-subjects factor and accuracy score for the test condition was the

dependent variable. Group differences were not observed on Non-overlapping Trained [F(2, 45)

= 1.71,p = .19, r12 = .07] or Non-overlapping Novel [F(2, 45) = 1.93, p = .16, r2 = .08]

conditions. A trend toward group differences was detected on the Overlapping Trained condition

[F(2, 45) = 2.68, p = .08, r12 = .11], with right ATL performing more poorly than controls. The

only test condition in which significant group differences were detected was the Overlapping

Inference condition [F(2, 45) = 4.96, p = .01, r2 = .18] (Figure 3-3). Bonferroni adjusted post-

hoc tests showed that patients with right ATL performed worse than controls, t(45) = 3.15, p=

.009. This finding supports the hypothesis that patients would have selective difficulty with

nonverbal TI after right ATL.









evaluating neurologically impaired populations. Secondly, on the test, 52% of participants

achieved greater than 90% correct on the TI condition. This ceiling effect suggests that a more

difficult task could yield greater information about the range of performance. Designing a more

difficult TI task could be easily accomplished by creating a larger hierarchy of overlapping pairs

that consists of more than five items. This would include more true TI pairs, like B>D, which do

not contain end items.

Also, condensing the task to focus only on the overlapping pairs may optimize clinical

utility and brevity. Results showed that conditions with overlapping pairs tended to be more

difficult, as evidenced by lower accuracy scores and slower reaction times. The paradigm in the

current study included conditions to demonstrate dissociations between transitive vs. non-

transitive inferential problems and between overlapping vs. non-overlapping pairs. This type of

paradigm may be more useful in neuroimaging studies, in which it is critical to design tasks that

highlight functional dissociations between neuroanatomical regions. In contrast, in a clinical

study, the information provided by the additional conditions with non-overlapping pairs is less

useful.

This study focused on a nonverbal task, given the dearth of behavioral measures sensitive

to language non-dominant hippocampal function. However, the study design did not permit a

full evaluation of material specificity. A verbal TI analogue would be needed to fully evaluate

the material-specific model of memory. The author developed a verbal TI task using non-words

as stimuli and piloted this task with healthy adults and with several epilepsy patients. Individuals

tended to perform more poorly on the verbal task and described it as more difficult. While the

nonverbal task may have permitted dual encoding resulting in better performance, the verbal task

may have facilitated strictly verbal encoding. Because of difficulties obtaining adequate pilot









Relationship between Transitive Inference and Standard Neuropsychological Tests

A series of partial correlations controlling for age examined relationships between the

experimental task and neuropsychological test performance (Table 3-6). Both non-TI (i.e.,

inferences about non-overlapping pairs) and TI (i.e., inferences about overlapping pairs) were

examined. All significant correlations were positive and ranged from small to moderate (r = .3

to .4). Non-TI was significantly correlated with two verbal memory measures (WMS-III Logical

Memory and Verbal Paired Associates) and several measures of attention/ executive functioning

(WAIS-III Digit Span, Trail Making Test Part B, and Wisconsin Card Sorting Test Errors). In

addition, non-TI and TI were correlated with the copy condition of the Rey Complex Figure

Test, a visuospatial, constructional measure. TI was significantly correlated with WAIS-III Full-

Scale IQ score, WMS-III Logical Memory (Total and Delayed), the Rey Complex Figure Test

(Immediate and Delayed), and Trail Making Test Part B.

Prediction of Right vs. Left Side of Resection

Binary logistic regression was conducted to evaluate neuropsychological predictors of side

of surgery. Two neuropsychological scores were entered as predictors, the Rey Complex Figure

Test Delayed Recall raw score and the TI score. Age was also specified as an a priori predictor.

The model achieved an overall classification accuracy of 73.9%, which is not a significant

improvement [2(3, 23) = 3.03, p = .39] over the base rate classification of 65%. While the

model correctly classified 93.3% of those who underwent right ATL, it correctly classified only

37.5% of those who underwent left ATL. Odds ratios (i.e., post-test probability of having

undergone right ATL) were calculated as .97 for the Rey Complex Figure score and 1.43 for the

TI score.

Receiver operating characteristic (ROC) curve analyses were performed to determine cut

scores that minimize diagnostic errors. The ROC curve analyses plot the rate of false positives









Cohen, N. J., Ryan, J., Hunt, C., Romine, L., Wszalek, T., & Nash, C. (1999). Hippocampal
system and declarative (relational) memory: Summarizing the data from functional
neuroimaging studies. Hippocampus, 9, 83-98.

Davachi, L., & Wagner, A. D. (2002). Hippocampal contributions to episodic encoding: Insights
from relational and item-based learning. Journal ofNeurophysiology, 88, 982-990.

Davis, R. N., Andresen, E. N., Witgert, M. E., & Breier, J. I. (2006). Is basic memory structure
invariant across epilepsy patient subgroups? Journal of Clinical andExperimental
Neuropsychology, 28, 987-997.

Devito, L. M., Kanter, B. R., & Eichenbaum, H. (in press). The hippocampus contributes to
memory expression during transitive inference in mice. Hippocampus.

Dusek, J. A., & Eichenbaum, H. (1997). The hippocampus and memory for orderly stimulus
relations. Proceedings of the NationalAcademy of Sciences USA, 94, 7109-7114.

Eichenbaum, H. (2000). A cortical-hippocampal system for declarative memory. Nature Reviews
Neuroscience, 1, 41-50.

Eichenbaum, H. (2004). Hippocampus: Cognitive processes and neural representations that
underlie declarative memory. Neuron, 44, 109-120.

Engel, J., Jr. (1996). Introduction to temporal lobe epilepsy. Epilepsy Research, 26, 141-150.

Engel, J., Jr. (2001). Mesial temporal lobe epilepsy: What have we learned? Neuroscientist, 7,
340-352.

Frank, J., & Landeira-Fernandez, J. (2008). Comparison between two scoring systems of the
Rey-Osterrieth Complex Figure in left and right temporal lobe epileptic patients. Archives
of Clinical Neuropsychology, 23, 839-845.

Fortin, N. J., Agster, K. L., & Eichenbaum, H. B. (2002). Critical role of the hippocampus in
memory for sequences of events. Nature Neuroscience, 5, 458-462.

Giovanello, K. S., Schnyer, D. M., & Verfaellie, M. (2004). A critical role for the anterior
hippocampus in relational memory: Evidence from an fMRI study comparing associative
and item recognition. Hippocampus, 14, 5-8.

Glosser, G., Cole, L., Khatri, U., DellaPietra, L., & Kaplan, E. (2002). Assessing nonverbal
memory with the Biber Figure Learning Test-Extended in temporal lobe epilepsy patients.
Archives of Clinical Neuropsychology, 17, 25-35.

Goodglass, H. & Kaplan, E. (2000). Boston Naming Test-II. Philadelphia: Lippincott, Williams,
& Wilkins.

Gooding, P. A., Mayes, A. R., & van Eijk, R. (2000). A meta-analysis of indirect memory tests
for novel material in organic amnesics. Neuropsychologia, 38, 666-676.









CHAPTER 4
DISCUSSION

Transitive Inference Performance in Temporal Lobe Epilepsy

This study examined performance on a transitive inference (TI) paradigm in patients who

had a history of intractable TLE and who had undergone standard ATL for seizure relief. During

training on the task, participants demonstrated learning of the visual stimulus pairs through

improved accuracy and faster reaction times as training progressed. Epilepsy patients achieved

similar levels of accuracy as controls. On the test, patients who had undergone right ATL

performed significantly worse than control participants, and patients who had undergone left

ATL performed in the intermediate range. Significant performance differences were not

observed between patients based on side of surgery.

Importantly, the right ATL group performed more poorly on the TI condition, which

involved making inferences across a series of overlapping pairs that formed a hierarchy. On the

TI condition, patients with right ATL achieved a mean score of only 71% correct, while patients

with left ATL and control participants achieved mean scores of 82% and 89% respectively. The

effect size for this group difference was in the medium range, and this difference remained

evident after controlling for the effects of age and initial learning. This finding supports the

primary hypothesis that patients with right ATL would have selective difficulty with TI for

patterned shapes. As expected, the true TI pair (BD), in which both stimuli had been equally

reinforced during training, was the most difficult individual pair for right ATL patients.

An examination of demographic and clinical variables showed that age was significantly

related to accuracy scores. Older individuals tended to perform more poorly, which is consistent

with prior research indicating that older adults experience difficulty with the relational

organization of propositions within memory, although previous studies had employed samples









epilepsy [/(1) = .89, p = .37], history of febrile seizures [2(1) = .50, p = .48], pre-surgical

seizure frequency [t(22) = .33, p = .75], seizure classification [/(1) = .27, p = .60], pre-surgical

MRI results [/(2) = 2.68, p = .26], time since surgery [t(22) = -1.34, p = .19], or post-surgical

seizure outcome [/2(2) = .66, p = .72]. Also, no patient had a history of vagal nerve stimulator

implantation.

Data Preparation

Score Derivation

Scores for accuracy and response latency were derived for performance on the transitive

inference (TI) task. Accuracy scores were generated by taking the mean percentage correct.

Latency scores were derived by first eliminating all response times greater than 15 seconds,

which likely reflected the influence of extraneous factors. Then, for each individual participant,

the median response time for correct responses was derived for each part of the task. This

resulted in the elimination of 3,765 data points out of 21,504. Accuracy and latency scores were

derived for training on each stimulus set (Block 1: Trials 1-48, Block 2: Trials 49-96, and Block

3: Trials 97-144) and for the test conditions: 1) Non-overlapping Trained, 2) Non-overlapping

Inference, 3) Overlapping Trained, and 4) Overlapping Inference.

Data Screening

The assumptions of univariate normality were checked for each of the dependent variables.

This preliminary examination showed that distributions for TI accuracy scores were mildly to

moderately negatively skewed. Therefore, square-root transformation was performed to improve

normality. In subsequent analyses, results using the square-root transformed accuracy data are

reported. The analyses were also conducted using the raw TI accuracy data, which yielded the

same pattern of results as the square-root transformed data. The TI reaction times scores showed

some evidence of positive skewness and elevated kurtosis for several variables. An outlier









Table 3-6. Partial correlations between transitive and non-transitive inference and performance
on standard neuropsychological tests after controlling for age
Non-transitive inference Transitive inference

r r

WASI Full-Scale IQ .26 .41**

HVLT-R
Immediate total -.10 .21
Delayed recall .11 .21

WMS-III Logical Memory
Immediate total .31* .32*
Delayed recall .35* .39**

WMS-III Verbal Paired Associates
Immediate total .30* .24
Delayed recall .07 .14

Rey Complex Figure Test
Copy (raw) .37* .33*
Immediate recall .14 .33*
Delayed recall .14 .32*
Delayed recognition .26 .27

Boston Naming Test-II .26 .28

WAIS-III Digit Span .40** .16

Trail Making Test
Part A -.04 .24
Part B .39** .38*

Wisconsin Card Sorting Test
Errors .34* .24
Perseverations .26 .22
Note. Square-root transformed accuracy scores represent performance on the TI task. Standard, scaled, or T scores
represent performance on standard neuropsychological measures unless otherwise noted. N=47. Missing data
included five Full-Scale IQ, one Digit Span, seven WCST, one Trail Making Test, three Boston Naming Test, two
Verbal Paired Associates Delayed, one Logical Memory, and one Rey Complex Figure Test. *p < .05. **p < .01.









hippocampal pathology that can be used to examine hippocampal contributions to memory.

Examining relational memory in TLE has the potential not only to provide support for relational

memory theory in a new population, but also to improve clinical care. The following sections

describe the role of neuropsychology in the clinical care of epilepsy and detail problems with

current clinical memory tests used to assess TLE patients. The key idea is that many of these

problems can be rectified through the development of more sensitive and specific tests of

hippocampal function.

The Role of Neuropsychological Testing in the Surgical Treatment of Epilepsy

Despite medical treatment, approximately 30 percent of the total epilepsy population

continues to experience seizures that are considered intractable (Helmstaedter, 2004). The

seizures associated with mesial TLE are among the most resistant to antiepileptic drugs (Engel,

2001). For treatment refractory TLE cases, surgery to remove a portion of the anterior temporal

lobes is an option, giving patients a 70 to 90 percent chance of becoming seizure free (Engel,

2001). The best surgical candidates show unilateral hippocampal sclerosis, ictal and interictal

abnormalities limited to relatively discrete neural zones, and focal/lateralized neuropsychological

deficits (Helmstaedter, 2004; Loring, 1997). The current standard of care for pre-surgical

candidates includes neuropsychological testing designed to yield information about localization

and lateralization of seizure focus and to help predict whether iatrogenic postoperative cognitive

disability will result from the planned resection (Helmstaedter, 2004). The degree of deviation

from expected pattern of performance on neuropsychological testing may serve as a prognostic

indicator of post-operative cognitive decline.

While recovery of cognitive function is often noted after surgery, there is also a risk for

cognitive decline. Estimates are that approximately 40 percent (range 10-60 percent) of patients

experience verbal memory loss after anterior temporal lobectomy (ATL; Chelune, 1995; Loring,









Test


During testing, memory for previously viewed pairs and the ability to infer the correct

response for novel pairings were assessed. No reinforcement was provided. Participants were

instructed:

In this section, you will see pairs from Part 1 and Part 2. Your job is to pick the correct
object. You will no longer see the smiley face, even if you pick the correct object. Pick the
object that you think would be correct based on what you learned in practice. In this part,
you will also see objects paired in new ways. When this happens, please make your best
guess about which object should be correct. Think about what you learned in practice and
about the objects in relation to their partners. The objects will be on the screen for a
limited time, so try to respond as quickly and as accurately as you can.

Two blocks of testing were administered, and each block consisted of 80 trials divided into

10 trials of a particular type (Trained Non-overlapping, Inference Non-overlapping, Trained

Overlapping, Inference Overlapping). Participants were asked to make inferences about five

novel pairings from the non-overlapping stimulus set and five novel pairings from the

overlapping stimulus set (Table 2-1). An inference from the non-overlapping set always paired

an object that had been a "winner" with an object that had never been reinforced. Therefore, the

problem could be solved by simply remembering the reinforcement history of each object. In

contrast, inferences from the overlapping set could pair objects that had not been consistently

reinforced during training (i.e., object had been "winner" and "loser" depending on its partner).

For instance, B>D is a critical pair, in which both objects were reinforced on 50% of training

trials. In order to solve this transitive problem, it would be essential to encode the overlapping

set as a hierarchy and to remember how a stimulus relates to other stimuli in the hierarchy.

Scoring and interpretation

Indices derived from performance included accuracy and latency scores. A key score was

accuracy (percentage correct) for the Overlapping Inference condition, which reflects the process

of TI. In the context of the current study, it was expected that patients who had undergone right









Scoville, W. B., & Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions.
Journal ofNeurology, Neurosurgery, & Psychiatry, 20, 11-21.

Snitz, B. E., Roman, D. D., & Beniak, T. E. (1996). Efficacy of the Continuous Visual Memory
Test in lateralizing temporal lobe dysfunction in chronic complex-partial epilepsy. Journal
of Clinical and Experimental Neuropsychology, 18, 747-754.

Sperling, R. A., Bates, J. F., Cocchiarella, A. J., Schacter, D. L., Rosen, B. R., & Albert, M. S.
(2001). Encoding novel face-name associations: A functional MRI study. Human Brain
Mapping, 14, 129-139.

Squire, L. R. (2004). Memory systems of the brain: A brief history and current perspective.
Neurobiology ofLearning and Memory, 82, 171-177.

Staresina, B. P., & Davachi, L. (2006). Differential encoding mechanisms for subsequent
associative recognition and free recall. The Journal ofNeuroscience, 26, 9162-9172.

Stark, C. E., Bayley, P. J., & Squire, L. R. (2002). Recognition memory for single items and for
associations is similarly impaired following damage to the hippocampal region. Learning
and Memory, 9, 238-242.

Stark, C. E., & Squire, L. R. (2003). Hippocampal damage equally impairs memory for single
items and memory for conjunctions. Hippocampus, 13, 281-292.

Strauss, E., Loring, D., Chelune, G., Hunter, M., Hermann, B., Perrine, K., et al. (1995).
Predicting cognitive impairment in epilepsy: Findings from the Bozeman epilepsy
consortium. Journal of Clinical and Experimental Neuropsychology, 17, 909-917.

Stroup, E., Langfitt, J., Berg, M., McDermott, M., Pilcher, W., & Como, P. (2003). Predicting
verbal memory decline following anterior temporal lobectomy (ATL). Neurology, 60,
1266-1273.

Titone, D., Ditman, T., Holzman, P. S., Eichenbaum, H., & Levy, D. L. (2004). Transitive
inference in schizophrenia: Impairments in relational memory organization. Schizophrenia
Research, 68, 235-247.

Trahan, D. E., & Larrabee, G. J. (1988). Continuous VisualMemory Test. Odessa, FL:
Psychological Assessment Resources.

Trenerry, M. R. (1996). Neuropsychologic assessment in surgical treatment of epilepsy. Mayo
Clinic Proceedings, 71, 1196-1200.

Trenerry, M. R., Westerveld, M., & Meador, K. J. (1995). MRI hippocampal volume and
neuropsychology in epilepsy surgery. Magnetic Resonance hiuging. 13, 1125-1132.

Turriziani, P., Fadda, L., Caltagirone, C., & Carlesimo, G. A. (2004). Recognition memory for
single items and for associations in amnesic patients. Neuropsychologia, 42, 426-433.









for the pair was the dependent variable. Accuracy for the pairs, AC, AD, and BE, did not differ

significantly between groups (p > .35). Group differences were observed on the BD pair [ 2 (2)

= 6.31, p = .04, r2= .13] and the CE pair [2(2) = 6.24,p = .04, rl= .13]. Post-hoc analyses,

using either the Mann Whitney Utest or rank sums test, showed that patients who underwent

right ATL performed more poorly than controls on BD (p < .05). Post-hoc tests revealed no

significant group differences on the CE pair.

Hierarchical Awareness

Following the computer-based task, it was assessed whether participants were aware of the

hierarchy in the overlapping stimulus set. First, they were presented with hand-held cards

representing B and D stimuli from the overlapping set and asked to select the winner. The

proportion of participants who correctly selected B was 56.2% of patients with right ATL, 75.0%

of patients with left ATL, and 87.5% of controls. A chi-square analysis revealed that these

differences were not statistically significant, 2 (2) = 5.00, p = .08, but showed a trend in the

predicted direction.

Participants were then presented with five hand-held cards representing the stimuli from

the overlapping set and asked to arrange them in order of dominance. A hierarchical awareness

score (range 0-12) was derived, with lower scores indicative of greater hierarchical awareness.

The hierarchical awareness score was related to performance on the Overlapping Inference

condition (r = .56, p < .001), after controlling for the influence of age and initial learning.

Awareness of the hierarchy accounted for 32% of the variance in scores on the Overlapping

Inference condition. Hierarchical awareness scores were calculated for each group, control (M=

3.1, SD = 2.9), left ATL (M= 3.7, SD = 4.6), and right ATL (M = 4.4, SD = 2.9). A one-way

ANOVA with group as the between-subjects factor and hierarchical awareness as the dependent

variable was performed. No group differences were detected, F(2, 45) = .76, p = .47, l = .03.









the Overlapping Inference test condition appeared most sensitive to right ATL. Therefore, this

'TI score' was selected to represent performance in the subsequent analyses for this aim.

Standard Neuropsychological Test Performance

Performance on standard neuropsychological tests was examined using a series of one-way

ANOVAs with group as the independent variable (right ATL, left ATL, control) and test score as

the dependent variable (Table 3-5). Groups were comparable on intellectual functioning, which

was in the average range.

As expected, significant group differences were observed on verbal memory tests,

including scores for HVLT-R Total [F(2, 44) = 7.63, p = .001, r2 = .26], HVLT-R Delayed

Recall [F(2, 44) = 6.51, p = .003, r2 = .23], WMS-III Logical Memory Total [F(2, 43) = 4.39, p

= .02, 12 = .17], WMS-III Verbal Paired Associates Total [F(2, 44) = 9.53,p < .001, 12 = .30],

and Verbal Paired Associates Delayed Recall [F(2, 42) = 5.49, p = .008, r12 = .21]. Group

differences were also observed on the Boston Naming Test-II [F(2, 41) = 7.51, p = .002, 2 =

.27]. Bonferroni-corrected post-hoc tests showed that patients with left ATL performed worse

than controls on these verbal memory scores and on the Boston Naming Test-II (p < .05).

Furthermore, the HVLT-R (Total and Delayed Recall) and the WMS-III Verbal Paired

Associates (Total) discriminated between side of surgery, such that patients with left ATL

performed significantly worse than those with right ATL (p < .05).

In contrast, group differences were not observed on a measure of nonverbal memory, the

Rey Complex Figure Test. Mean scores for immediate and delayed recall on this measure

ranged from the 5th to the 20th percentile across groups. Also, group differences were not

detected on measures of attention or executive function.











1900

1700

u 1500
E
) 1300
E
" 1100

E 900

700

500


Sto 48


49 to 96

Training block


-- RATL
-- -LATL
--- Control


97 to 144


Figure 3-5. Reaction time for non-overlapping pairs by training block.


-+-RATL
-- -LATL
--- Control


1to48 49 to 96 97 to 144

Training block


Figure 3-6. Reaction time for overlapping pairs by training block.


~--9


1900

1700

, 1500
E
0) 1300
E
= 1100

c 900

700

500









group and training block [F(3.23, 72.62) = .56, p = .66, r2 = .02] was significant. Given that the

power to detect group differences was only .45, exploratory one-way ANOVAs were conducted

to further examine performance on each block. Group differences were detected only on Block 1

of the task, F(2, 45) = 4.34, p = .02, r2 = .16. Patients with right ATL were found to perform

more poorly than controls on Block 1, t(45) = 2.73, p = .03.

Training performance on overlapping pairs was also evaluated (Figure 3-2). The main

effect of training was significant, F(2, 90) = 19.97, p < .001, r1 = .31. Bonferroni adjusted post-

hoc tests showed that participants improved across training from Block 1 to Block 2 (p = .001)

and from Block 1 to Block 3 (p < .001). A trend for improvement from Block 2 to Block 3 (p =

.08) was observed. Neither the main effect of group [F(2, 45) = 2.17, p = .12, rf2 = .09], nor the

interaction between group and training block [F(4, 90) = 1.97, p = .10, r12 = .08] was significant.

Given that the power to detect differences between groups was only .42, exploratory one-way

ANOVAs were conducted to further examine performance on each block. Group differences

were detected only on Block 2 of the task, F(2, 45) = 3.90, p = .03, r2 = .15. Patients with left

ATL performed more poorly than controls on Block 2, t(45) = 2.53, p = .04.

Taken together, these findings demonstrate that participants improved across training for

both the non-overlapping and overlapping sets. Each group achieved mean accuracy of at least

85% on the third training block for non-overlapping pairs and at least 78% on the third training

block for overlapping pairs. Groups achieved fairly comparable levels of learning for the non-

overlapping and the overlapping stimulus sets by the third training block.

Accuracy: Test

Test performance was evaluated using a 3 (Group: right ATL, left ATL, control) x 4

(Condition: Non-overlapping Trained, Non-overlapping Inference, Overlapping Trained,

Overlapping Inference) mixed between-within ANOVA. For the within-subjects factor,










3 R E S U L T S ................... .......................................................... ................ 3 6

Demographic and Clinical Characteristics of Participants.............................................36
D ata P re p a ra tio n ............................................................................................................... 3 7
Score D erivation ....................................................... ........ ................. .. 37
Data Screening............................................. 37
Aim 1: Transitive Inference Task Performance................................... ....................... 38
A accuracy: Training ....................................................... .......... ...... .... ...38
A ccu racy : T est.......................................................39
H ierarchical A w areness.......... .............................................................. .. .... ..... .. 43
R action T im e: T raining ........... ............................................................ .. .... ...... 44
R action T im e: T est .................... ........ ...................................... .. ........... .... 45
Aim 2: Comparison of Transitive Inference and Standard Neuropsychological Tests ..........46
Standard Neuropsychological Test Performance ...................................................... 47
Relationship between Transitive Inference and Standard Neuropsychological Tests.....48
Prediction of Right vs. Left Side of Resection....................................................48

4 D ISC U S SIO N ..............................................................................................60

Transitive Inference Performance in Temporal Lobe Epilepsy............... ...............60
Transitive Inference and Cognitive Neuroscience Research............................. ...............62
Clinical Applications of Transitive Inference..................................... ......... ............... 64
Lim stations and Future D directions ............ ................................. .................. ............... 66
C on clu sion .............. ....................................................... ...........................6 9

L IST O F R E F E R E N C E S ......... .. ................. ..........................................................................70

B IO G R A PH IC A L SK E T C H ............................................ ......... ................... ...........................77























6









CHAPTER 3
RESULTS

Data were analyzed using the SPSS 16 statistical software package. An alpha level of .05

was used throughout the analyses.

Demographic and Clinical Characteristics of Participants

Descriptive statistics were generated for the total sample and the groups (right ATL, left

ATL, control) (Table 3-1). The groups were contrasted on demographic and clinical variables

using one-way analyses of variance (ANOVAs), independent samples t-tests, and Pearson chi-

square tests. Participants ranged in age from 20 to 66 years (M= 41.8, SD = 14.0). Educational

level ranged from nine to 20 years (M= 14.9, SD = 2.8). Gender composition of the sample was

evenly balanced, with 50% female. The sample was predominantly Caucasian (83%), with a

smaller proportion of African-Americans (13%) and persons of Hispanic origin (4%). With the

exception of one patient, all participants reported primarily right-hand dominance as assessed by

the Edinburgh Handedness Inventory. No group differences were detected on demographic

characteristics including age [F(2, 18.7)= .54, p = .59], education [F(2, 44) = 2.80, p = .07],

gender [V(2) = .42, p = .81], or race/ethnicity [/(4) = 3.83, p = .43].

The epilepsy groups were well matched on disease characteristics (Table 3-1). Mean age

at onset of epilepsy was 20.5 years (SD = 14.1), and the mean duration of epilepsy was 19.6

years (SD = 14.0). 41.7% of patients had experienced a mild to moderate traumatic brain injury,

which may have been a risk factor in the development of their epilepsy. The time from surgery

to study evaluation ranged from eight to 95 months, with a mean of 55 months (SD = 26.8).

Regarding surgery outcome, 66.7% of patients were classified as seizure and aura free after

surgery. No group differences were detected on clinical variables, including age at onset of

epilepsy [t(22) = -.77, p = .45], duration of epilepsy [t(22) = 1.10, p = .28], family history of









4) Rey Complex Figure Test and Recognition Trial (RCFT; Meyers & Meyers, 1995). The

RCFT assesses visuospatial constructional ability and nonverbal memory. The test involves

copying a complex geometric design, which the participant then draws from memory after a 3-

minute delay and after a 30-minute delay. A recognition trial follows the delayed free recall.

The drawings are scored based on the accuracy and placement of the elements of the design.

5) Boston Naming Test-II (BNT-II; Goodglass & Kaplan, 2000).* This measure of

confrontation naming includes 60 black-and-white ink drawings. Participants are asked to name

each picture, and semantic and phonemic cues can be provided.

6) Digit Span subtest from the Wechsler Adult Intelligence Scale-Third Edition (WAIS-III;

Wechsler, 1997a). On the Digit Span subtest, examinees are read a string of digits and asked to

recall them in order. The string increases by one after two trials are completed correctly. Then,

using the same procedure, the ability to recall digit strings in reverse order is tested.

7) Trail Making Test (Army Individual Test Battery, 1944; Reitan, 1958).* This test

measures visuomotor speed, attention, tracking, and set-shifting. In Part A, participants are

required to connect numbers from 1 to 25 and instructed to do so as quickly as possible. In Part

B, participants are asked to connect letters and numbers in order, shifting between sets.

8) Wisconsin Card Sorting Test (WCST; Heaton, 1981).* This measure of mental

flexibility and problem solving requires examinees to sort two decks of 64 cards with colored

shapes printed on them. The cards should be sorted based on unstated principles, such as color,

form, or number.

9) Edinburgh Handedness Inventory (Oldfield, 1971). This ten-item questionnaire assesses

hand preference for a range of activities. A laterality quotient ranging from -1 (left-hand

dominance) to +1 (right-hand dominance) can be calculated.




























49 to 96


--RATL
--C--LATL
-A- Control


97 to 144


Training block


Figure 3-1. Accuracy for non-overlapping pairs by training block.


49 to 96


-- RATL
- o--LATL
-A- Control


97 to 144


Training block


Figure 3-2. Accuracy for overlapping pairs by training block.


1 to 48


1 to 48


Ar









The Hippocampus: The Physiological Basis for Relational Memory

According to relational memory theory, the hippocampus subserves relational mechanisms

important for binding together the cognitive, affective, and contextual features of a learning

event into an integrated memory (Eichenbaum, 2004). The hippocampus is uniquely positioned

to serve as a relational binding mechanism (Eichenbaum, 2000). Information from association

cortices converges on the parahippocampal and perirhinal cortices, which in turn project to the

entorhinal cortex (Preston & Gabrieli, 2002). The entorhinal cortex serves as the primary input

mechanism to the hippocampus and provides segregated input from a wide range of cortical

areas.

The placement of the hippocampus as the source of converging cortical input and its

capacity to form associations make it well equipped to serve as the site of relational memory.

Although input to the hippocampus is segregated with regard to sensory modality or point of

origin, the structure of the hippocampus itself is thought to lend itself to associational processing.

Within the hippocampus, information proceeds in an orderly, unidirectional manner. O'Reilly

and Rudy (2001) developed a biologically-based computational model of hippocampal function

and proposed that relational binding occurs in a rapid, automatic manner. Integral to this model

are the auto-associative binding features of CA3, an area in the hippocampus (Eichenbaum,

2004; O'Reilly & Rudy, 2001). Hippocampal properties that are prominent in the CA3 region

and contribute to the capacity to form associations include broad recurrent connections and the

prevalence of rapid synaptic plasticity, known as long-term potentiation (Eichenbaum, 2004).

The physiological basis of the proposed architecture to support relational memory processing is

one of the strengths of the relational memory concept.

While the hippocampal formation is necessary for processing relations among multiple

stimuli, an emerging idea is that surrounding cortices mediate performance on simpler tasks that









may be more sensitive to left TLE. This pattern would be predicted by relational memory

theory. The WMS-III Logical Memory subtest did not demonstrate the capability to discriminate

between patients based on side of surgery, although patients with left ATL did perform worse

than controls on this task.

Conversely, the conventional nonverbal memory measure did not discriminate between

groups (right ATL, left ATL, control). This finding is consistent with previous research, which

has not found a reliable link between right TLE and nonverbal memory (Barr et al., 1997;

Helmstaedter, 2004; Lee et al., 2002; Martin et al., 1999). Of note, the epilepsy patient groups

and the control group performed in the impaired to low average range on the Rey Complex

Figure Test. The low scores may be indicative of the false lateralizing figural memory

performance sometimes observed in left TLE (Loring et al., 2007).

The TI task may have more promise as a behavioral indicator of right hippocampal

dysfunction than the Rey Complex Figure Test. The motor and constructional abilities involved

in figural reproduction may confound the assessment of memory, while the TI task uses a

recognition format to bypass this concern. Furthermore, analysis revealed significant group

differences on TI (right ATL < control) with a medium effect size (r2 = .18 to .20). In contrast,

the Rey Complex Figure Test did not detect significant group differences, and a small effect size

(r2 = .03) was noted for both the immediate and delayed recall conditions of this measure.

The clinical utility of these measures was further explored using logistic regression and

receiver operating characteristic (ROC) curve analysis. Neither measure was a significant

predictor of side of surgery, and operating characteristics were similar for both measures. Low

cut scores are suggested: 75% for the TI score and 30 for the Rey Complex Figure Test Delayed

Recall T score. The cut scores were set to optimize positive predictive power and specificity,









Table 3-5. Neuropsychological test performance by group
Right ATL Left ATL Control F p
(n=16)a (n=8)b (n=24)c


WASI Full-Scale IQ (SS)
HVLT-R
Immediate total (T)
Delayed recall (T)
WMS-III Logical Memory
Immediate total (ss)
Delayed recall (ss)
WMS-III Verbal Paired As.
Immediate total (ss)
Delayed recall (ss)

Rey Complex Figure Test
Copy (raw)
Immediate recall (T)
Delayed recall (T)
Recognition (T)
Boston Naming Test-II (T)
WAIS-III Digit Span (ss)
Trail Making Test
Part A (T)
Part B (T)
Wisconsin Card Sorting Test
Errors (T)
Perseverations (T)


101.5 (13.0)


41.7 (10.2)
42.4(13.6)

9.4(2.1)
10.1 (2.6)


10.5 (2.1)
10.9 (2.6)


29.7 (4.8)
34.9(13.6)
34.1 (12.7)
41.7(10.1)
43.2(9.8)
10.9 (3.3)

39.9 (12.3)
47.1 (8.5)


45.2(10.9)
45.1 (10.3)


102.1 (9.9)


30.0 (9.4)
26.1 (8.0)

8.4(3.4)
9.7 (2.6)


6.9(3.1)
8.3 (3.8)


31.8(2.2)
41.7(9.2)
39.9 (9.3)
43.7(5.3)
34.1 (4.1)
9.9 (4.0)

46.4(11.2)
50.0(9.3)


50.4(10.6)
53.3 (12.7)


104.4 (10.8)


44.3 (7.8)
43.2(11.8)

11.3 (2.8)
11.6(2.4)


11.1 (2.3)
11.8(1.9)


31.0(4.2)
38.3 (15.0)
37.6(12.0)
48.4(11.5)
51.3 (13.3)
11.6(3.2)

43.9(10.4)
47.9(12.3)


50.6 (7.6)
51.4 (8.7)


7.63
6.51

4.39
2.50


9.53
5.49


.81
.67
.70
2.03
7.51


.02
.09


<.001
.008


.75 .48


1.03
.19


1.38
2.04


Note. SS=Standard Score, T=T score, ss=scaled score. aMissing data for right ATL included two Full-Scale IQ, one
Digit Span, four WCST, one Trail Making Test, three Boston Naming Test, one Verbal Paired Associates Delayed,
one Logical Memory, and one Rey Complex Figure Test. bMissing data for left ATL included one Full-Scale IQ,
one WCST, and one Verbal Paired Associates Delayed. cOne control was only administered experimental measures.
Two controls were missing the Full-Scale IQ, and two controls were not administered the WCST.




Full Text

PAGE 1

1 TRANSITIVE INFERENCE IN TEMPORAL LOBE EPILEPSY By MARIE D. BARKER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Marie D. Barker

PAGE 3

3 To my family

PAGE 4

4 ACKNOWLEDGMENTS I thank my dissertation chair, Dr. Russell Ba uer, for being a grea t mentor during my graduate training. I am very grateful for the exceptional guidance and support that he provided through all aspects of this project, from developing the concept to data collection and analysis. I would also like to express my sincere gratitude to the other member s of my dissertation committee, Michael Marsiske, David Loring, Eileen Fennell, and Steven R oper. I especially thank Dr. Michael Marsiske for his invaluable help with statistical issues and his insight into working with my experimental data. Lastly, I am grateful for the love and support of my wonderful husband and family throughout my graduate education.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4 LIST OF TABLES ...........................................................................................................................7 LIST OF FIGURES .........................................................................................................................8 ABSTRACT ...................................................................................................................... ...............9 CHAPTER 1 BACKGROUND AND SIGNIFICANCE .............................................................................. 11 Relational Memory .................................................................................................................11 The Hippocampus: The Physiological Basis for Relational Memory ............................. 12 Paradigms to Assess Relational Memory ........................................................................ 13 Paired associate learning .......................................................................................... 13 Transitive inference ..................................................................................................14 Converging Evidence for the Role of the Hippocampus in Relational Memory ............ 15 Animal models .........................................................................................................15 Neuroimaging of healthy adults ............................................................................... 16 Studies of amnesia .................................................................................................... 18 Relational Memory: Application to Temporal Lobe Epilepsy ...............................................19 The Role of Neuropsychological Testing in the Surgical Treatment of Epilepsy ........... 20 Assessment of Memory Functioning in TLE ..................................................................21 Improving Detection of Memory Dysfunction in Right TLE ......................................... 23 Purpose of Current Study ........................................................................................................25 2 METHODS ....................................................................................................................... ......26 Participants .................................................................................................................. ...........26 Procedures .................................................................................................................... ...........26 Recruitment Strategy ....................................................................................................... 26 Assessment Procedures ................................................................................................... 27 Measures ...................................................................................................................... ...........28 Transitive Inference Task ................................................................................................ 28 Training .................................................................................................................... 28 Test .......................................................................................................................... .30 Scoring and interpretation ........................................................................................ 30 Post-test assessment ................................................................................................. 31 Standard Neuropsychological Tests ................................................................................32

PAGE 6

6 3 RESULTS ....................................................................................................................... ........36 Demographic and Clinical Charac teristics of Participants .....................................................36 Data Preparation .............................................................................................................. .......37 Score Derivation ..............................................................................................................37 Data Screening .................................................................................................................37 Aim 1: Transitive Inference Task Performance ...................................................................... 38 Accuracy: Training ..........................................................................................................38 Accuracy: Test ................................................................................................................ .39 Hierarchical Awareness ...................................................................................................43 Reaction Time: Training .................................................................................................. 44 Reaction Time: Test ........................................................................................................ 45 Aim 2: Comparison of Transitive Inference and Standard Neuropsychological Tests ..........46 Standard Neuropsychological Test Performance ............................................................ 47 Relationship between Transitive Inferen ce and Standard Neuropsychological Tests ..... 48 Prediction of Right vs. Left Side of Resection ................................................................ 48 4 DISCUSSION .................................................................................................................... .....60 Transitive Inference Performance in Temporal Lobe Epilepsy .............................................. 60 Transitive Inference and Cognitive Neuroscience Research .................................................. 62 Clinical Applications of Transitive Inference ......................................................................... 64 Limitations and Future Directions .......................................................................................... 66 Conclusion .................................................................................................................... ..........69 LIST OF REFERENCES ...............................................................................................................70 BIOGRAPHICAL SKETCH .........................................................................................................77

PAGE 7

7 LIST OF TABLES Table page 2-1 Protocol for the transitive inference task ........................................................................... 34 3-1 Demographic and clinical characteristics by group ........................................................... 51 3-2 Transitive inference accuracy scores as percent correct by group ..................................... 52 3-3 Correlations between TI test, demographic characteristics, and clinical variables ........... 55 3-4 Transitive inference reaction times by group .....................................................................55 3-5 Neuropsychological te st performance by group ................................................................58 3-6 Partial correlations between transitive and non-transitive inference and performance on standard neuropsychological te sts after controlling for age ......................................... 59

PAGE 8

8 LIST OF FIGURES Figure page 2-1 Transitive inference stimuli .............................................................................................. .35 3-1 Accuracy for non-overlapping pairs by training block ...................................................... 53 3-2 Accuracy for overlapping pairs by training block ............................................................. 53 3-3 Accuracy by test condition ................................................................................................ .54 3-4 Accuracy for overlapping pair s requiring transitive inference .......................................... 54 3-5 Reaction time for non-overlap ping pairs by training block ...............................................56 3-6 Reaction time for overla pping pairs by training block ...................................................... 56 3-7 Reaction time by test condition.......................................................................................... 57 3-8 ROC curves for the prediction of ri ght anterior temporal lobectomy. ............................... 57

PAGE 9

9 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TRANSITIVE INFERENCE IN TEMPORAL LOBE EPILEPSY By Marie D. Barker August 2009 Chair: Russell M. Bauer Major: Psychology Recent findings in cognitive neuroscience reveal that transitive inference (TI) tasks, which require the formation and recognition of stimul us associations across experiences, have good specificity in the measurement of hippocampa l functioning. Extant re search has focused on animals and healthy adults. This study is the fi rst to apply the TI para digm in temporal lobe epilepsy (TLE), which is a syndrome that provides a model to study hippocampal contributions to memory. Primary aims were 1) to examine TI performance and relationship to side of surgery in TLE and 2) to compare the clinical utility of th e TI task to standard neuropsychological tests. Participants included 24 pa tients with TLE, who had undergone anterior temporal lobectomy (ATL; left n=8, right n=16), and 24 h ealthy controls. They completed a computerbased TI task, which was adapted from a paradi gm that has demonstrated selective right hippocampal activation in functional imaging st udies (Heckers et al., 2004). During training, participants view pairs of patterned shapes and learn the winner in each pair (e.g., A>B, B>C). They are tested on their ability to recollect the correct response fo r previously seen pairs and to make inferences about novel pairings (e.g., A>C). The critical condition involves making inferences across a series of overlapping pa irs that form a hierarchy (A>B>C>D>E).

PAGE 10

10 On the test, patients who had undergone righ t ATL performed significantly worse than healthy controls on TI for visual information. Left ATL patients performed in the intermediate range; however, the task did not di scriminate between patients based on side of surgery. Results provide some evidence of a laterality effect and suggest that TI may be sensitive to hippocampally-mediated memory function. There is a clear need for better neuropsychological measures to assess language non-dominant (usually right) temporal lobe function, given the poor sensitivity and specificity of current tests. In this study, the conve ntional nonverbal memory measure, the Rey Complex Figure Test, did not di scriminate between groups. The TI task and the conventional nonverbal measure yielded simila r operating characteristics with good positive predictive power but poor sensi tivity. The TI task showed m odest clinical promise, and modifications that may improve its clinical utility are suggested.

PAGE 11

11 CHAPTER 1 BACKGROUND AND SIGNIFICANCE Relational Memory Relational m emory refers to the capacity to cr eate a flexible and integrated representation of an experience that mediates associations am ong elements of the experience (Cohen, Poldrack, & Eichenbaum, 1997). Both the individual elements as well as their larger structure are encoded in relational memory. Anatomic data suggest that the hippocampus is essential for relational binding, linking multiple inputs together to represent their relationships and to code overlapping features across different e xperiences (Cohen et al., 1999). The concept of relational memory fits into the current framework of multiple memory systems. This framework arose out of res earch on amnesia, which first implicated the hippocampus in memory (Scoville & Milner, 195 7). The current framework focuses on a distinction between declarative and nondeclar ative memory processing (Squire, 2004). Declarative memory is characterized by the cons cious recollection of facts and events, while nondeclarative memory does not involve conscious recollection and is demonstrated through facilitation of performance rather than recollection. Given that relational memory involves the medi ation of associations between information, it can be integral for both declarative and nondeclarative memory. Fo r example, relational memory is essential for a form of declarative memory that involves representing episodes. In episodic memory, an individual encodes how differen t elements relate to form a representation of a complex event (e.g., a party, a final examination) what is unique about the event, and how it links to other episodic memories through common elements (Eichenbaum, 2000; OReilly & Rudy, 2001).

PAGE 12

12 The Hippocampus: The Physiological Basis for Relational Memory According to relational me mory theory, the hippocampus subserves relational mechanisms important for binding together the cognitive, aff ective, and contextual features of a learning event into an integrated memory (Eichenbaum 2004). The hippocampus is uniquely positioned to serve as a relational binding mechanism (Eic henbaum, 2000). Information from association cortices converges on the parahippocampal and perirh inal cortices, which in turn project to the entorhinal cortex (Preston & Gabrieli, 2002). The entorhinal cortex serves as the primary input mechanism to the hippocampus and provides segreg ated input from a wide range of cortical areas. The placement of the hippocampus as the sour ce of converging cortical input and its capacity to form associations make it well equipped to serve as the site of relational memory. Although input to the hippocampus is segregated with regard to sensory modality or point of origin, the structure of the hippocampus itself is thought to lend itself to a ssociational processing. Within the hippocampus, information proceeds in an orderly, unidirectional manner. OReilly and Rudy (2001) developed a biologically-based computational model of hippocampal function and proposed that relational bindi ng occurs in a rapid, automatic ma nner. Integral to this model are the auto-associativ e binding features of CA3, an ar ea in the hippocampus (Eichenbaum, 2004; OReilly & Rudy, 2001). Hippocampal propert ies that are prominent in the CA3 region and contribute to the capacity to form associations include broa d recurrent connections and the prevalence of rapid synaptic plasticity, known as long-term potentiation (Eichenbaum, 2004). The physiological basis of the pr oposed architecture to support re lational memory processing is one of the strengths of the relational memory concept. While the hippocampal formation is necessa ry for processing re lations among multiple stimuli, an emerging idea is that surrounding cort ices mediate performance on simpler tasks that

PAGE 13

13 rely on stimulus novelty or familiarity (Preston & Gabrieli, 2002). Medial temporal structures implicated in more recognition-based item me mory include the perirhinal cortex (Aggleton & Brown, 1999; Brown & Aggleton, 2001; Koehle r, Danckert, Gati, & Menon, 2005) and entorhinal and parahippocampal cortices (Davachi & Wagner, 2002). The distinction between item and relational memory is subtle but impor tant, and focuses on whether the stimulus has been encoded as an item, which is a unitized whole, or as a relational representation, which preserves the constituent elements as well as the larger structure. It is important to recognize that most clinical memory tests are not designe d to be sensitive to this distinction. Paradigms to Assess Relational Memory Several paradigms have been developed to spec if ically assess relati onal memory. Two of the most widely used paradigms, paired associ ate learning and transiti ve inference (TI), both incorporate rapid, incidental learning. Paired associate learning One paradigm to assess relational me mory involves learning pair ed associates (e.g., Giovanello, Schnyer, & Verfaellie, 2004; Meltzer & Constable, 2005) The subject is trained on a series of pairs, A-B, C-D, E-F, where A-F can be words or pictures. During a test phase, the subject is presented with intact pairs (e.g., AB), novel pairs (e.g., Y-Z), and recombined pairs (e.g., A-F; items presented previously, but not together). Due to their similar exposure histories, items in intact and recombined pairs are equally familiar to subjects and so differ only in whether they were previously associated. Performance differences on intact versus recombined pairs provides an index of relational memory, while performance differences on recombined versus novel pairs provides a measure of item memory. Thus, the paired associ ate paradigm allows unique assessment of relational and item memory and eliminates the confound of familiarity on performance. This paradigm has demonstrated sensitivity to hippocampal function in amnesic

PAGE 14

14 patients with medial temporal lobe damage (Kroll, Knight, Metcalfe, Wolf, & Tulving, 1996) and in functional imaging studies (Giovanello et al., 2004; Meltzer & Constable, 2005). Transitive inference Another approach to test relational memory i nvolves training subjects with distinct stimuli that sh are common elements and then testing whether these experiences have been linked in memory to solve new problems (Eichenbaum, 2000) In the TI paradigm, subjects are exposed to a series of overlapping stimulus pairs (e.g., A>B, B>C, where > mean s is preferred to in regards to the likelihood of obtaining reinforcement when that ite m is selected over the other). A measure of relational memory is provided by the subjects ability to make an inference about stimuli that were not previously presented togeth er, but are related throug h the overlapping pairs. For example, TI would be demonstrated by the k nowledge that A>C. The overlapping pairs are likely stored as a flexible representation that can be manipulated mentally to solve the transitive problem (Heckers, Zalesak, Weiss, Ditman, & Tit one, 2004). Although this effect is episodic in the sense that it arises out of exposure to the stimuli, it is i nferential in that it involves reference to a memory representation that is not based on direct experience, since during learning, A and C were never presented together. The TI paradigm is based on a well-validated animal model, and the capacity for TI has been demonstrated in rodents (Dusek & Ei chenbaum, 1997), pigeons (von Fersen, Wynne, Delius, & Staddon, 1991), and nonhuman primates (McGonigle & Chalmers, 1977) that were trained in instrumental learning fo rmats. This paradigm has also been applied to humans in a comparative approach. In healthy adults, TI has primarily been tested using visual materials, such as geometric designs (Heckers et al., 2004, Heckers & Zalesak, 2009), faces (Nagode & Pardo, 2002), and face-house pairings (Preston, Shrager, Dudukovic, & Gabrieli, 2004).

PAGE 15

15 TI was selected as the primary measure for th e current study because it appears to uniquely capture the role of the hippocampus in encodi ng relationships among multiple stimuli, rather than simply encoding relationships between two simultaneously presented items as in paired associates. It extends the concept of paired associates to evaluate relationships across pairs. In addition, TI may be better representative of daily demands on memory processing, as this paradigm highlights the application of memory in a novel situation. The ability to utilize information learned across situations, to make in ferences, and to apply that knowledge in a novel situation is critical for higher cognitive function. Converging Evidence for the Role of th e Hippocampus in Relational Memory The concept of relational m emory adds signi ficantly to our unde rstanding of how the hippocampus is involved in memory. Studies u tilizing the paradigms described in the preceding section have provided strong support for the relational memory account in the animal, neuroimaging, and neuropsychological literatures. Animal models In a review of animal studies, Eichenbaum ( 2000) reported that defi cits in well-validated tasks used in the animal memory literature, such as a water maze and TI, reflect problems in relational memory. Several studies have demonstrat ed relational memory deficits in rodents with hippocampal lesions. For instance, Fortin, Ag ster, and Eichenbaum (2002) showed that hippocampally-lesioned rats were impaired on lear ning the sequential order of a series of odors, but not on recognition of previously learned odo rs. Similarly, Dusek and Eichenbaum (1997) found that hippocampal disconnections from cor tical and subcortical pathways selectively impaired TI in odor learning. Later studies have attempted to pinpoint the role of the hippocampus. The hippocampus may help to ac quire the underlying repr esentations necessary for TI in rodents (Van der Jeugd et al., 2009) On a different note, Devito, Kanter, and

PAGE 16

16 Eichenbaum (in press) showed that hippocam pal damage produced after mice learned an overlapping sequence did not substantially affect original learning but resulted in severe impairment in subsequent TI. This finding implies that the hippocampus is important for accessing the representation in a way necessary to perform TI. Neuroimaging of healthy adults Neuroima ging studies of healthy adults also provide evidence of the role of the hippocampus in relational memory. In a review of the functional imagi ng literature, Cohen and colleagues (1999) concluded that th e relational memory concept prov ided a better explanation of the data than other accounts of hippocampal functioning, which focus on novelty, the explicitimplicit memory distinction, or spatial mapping. The functional imaging studies typically rely on a difference analysis, such that the pattern of activation observed durin g an item memory task or other control task is compared to activa tion observed during a rela tional memory task. Areas that are selectively activated during the relational memory task, in comparison to the control task, are considered to reflect a unique contribution to relational memory. Functional imaging studies indicate that the framework for relational memory is laid down during the encoding process. For instance, greater hippocampal activa tion was observed during training for overlapping premise pairs that permitte d TI compared to those that did not (Nagode & Pardo, 2002). Similarly, other studies have demonstrated hippocam pal involvement during encoding of novel associations involving visual information (R ombouts et al., 1997; Sperling et al., 2001). Later studies have e xpanded on this work to show th at the extent of hippocampal activation during relational encoding correlates w ith performance on subsequent memory tasks. Hippocampal activity is greater dur ing encoding of relational info rmation that is successfully remembered on subsequent recall and recogni tion tasks (Davachi & Wa gner, 2002; Jackson & Schacter, 2004; Staresina & Davachi, 2006).

PAGE 17

17 While the majority of studies have focused on encoding processes, several have examined retrieval processes. Encoding a nd recall of associative memories are functions of an integrated hippocampal system (Meltzer & Constable, 2005). Research has revealed hippocampal activation during recognition of re lational information, bu t not item information (Preston et al., 2004; Yonelinas, Hopfinger, Buonocore, Kroll, & Baynes, 2001). Furthermore, some studies support the idea that relational en coding is associated with ante rior hippocampal activity, while retrieval is associated with posterior hippocampal activity (Meltz er & Constable, 2005; Prince, Daselaar, and Cabeza; 2005). In both studies, th e overlap of encoding and retrieval effects was maximal in the middle of the longitudinal ex tent of the hippocampus, near the CA3 area. Neuroimaging studies also provide evidence of selective hippocampal activation, relative to other temporal regions, during TI tasks (Nag ode & Pardo, 2002; Preston et al., 2004) and other relational memory tasks (Koehler et al., 2005). For instance, Heckers and colleagues (2004) identified a distributed network of brain regions, including th e pre-supplementary motor area, bilateral frontal cortex, bilateral parietal cortex, bilateral posterior tempor al cortex, and pulvinar, involved in TI for overlapping visual (nonverbal) stimulus pairs. Importantly, difference analysis demonstrated selective right anterior hippocampal activation du ring TI and bilateral activation in the anterior para hippocampal gyrus duri ng other task conditions. Moreover, in a similar study (Zalesak & Heck ers, 2009), greater hippocampal ac tivation was associated with more cognitively demanding aspects of TI. Speci fically, greater right hippocampal activation was observed for inferences about pairs derived fr om more adjacent items in the hierarchy, while greater left hippocampal activation was associated with inference pairs that did not contain end items versus those that did.

PAGE 18

18 Studies of amnesia Neuropsychological evidence also supports th e relational memory account. Patients with amnesia due to hippocampal damage have s hown selective difficulty with memory for associations (Kroll et al., 1996) and for conf igural learning, similar to the childhood rock, paper, scissors game (Rickard & Grafman, 1998). While a few studies have failed to replicate these results (Stark, Bayley, & Squire, 2002; Stark & Squire, 2003), potentially due to characteristics of the medial tem poral lobe damage in their participants, these findings have been confirmed in a number of studies. For instance, impaired between-item asso ciations (e.g., face-face, face-word) with preserved single item recognition has been observed in six patients with hippocampal lesions (Turriziani, Fadda, Caltagirone & Carlesimo, 2004). Likewise Mayes and colleagues (2004) reported that a patient with bilateral hippocam pal lesions was impaired on memory for crossmodal associations (e.g., object-l ocation, face-name), but not memory for individual items or intra-item associations. Hippocampal amnesics experience problems remembering relations among items, even with very short delays (Hannula, Tranel, & Cohen, 2006; Olson, Page, Moore, Chatterjee, & Verfaellie, 2006). The impairment in relational memory is not confined to declarative memory tasks and extends into the nondeclarative domain as well. Priming, a measure of nondeclarative memory, is demonstrated when previous exposure to a stimulus leads to a faster reaction time or improved accuracy when the stimulus is shown again. Investigators have observed impaired priming for new associations in patients with amnesia. For instance, amnesics with focal medial temporal lobe lesions demonstrated impaired priming bot h for associations between two words and for associations between two featur es of a single stimulus, but showed preserved item priming (Yang et al., 2003). In a meta-analysis on implic it memory in organic amnesia, amnesics were

PAGE 19

19 found to perform equivalently to controls on indi rect tests of familiar information, but worse on indirect tests for novel item and associativ e information (Gooding, Mayes, & van Eijk, 2000). Relational Memory: Applicatio n to Temporal Lobe Epilepsy The previous section showed how relational me mory is characterized by the for mation of a flexible, integrated represen tation that relies on hippocampal functioning. Strengths of the relational memory concept include a wellgrounded physiological basis and converging empirical support from studies of animals, health y adults, and clinical populations. To date, there have been no studies of relational memory in temporal lobe epilepsy (TLE) using the experimental paradigms describe d in the preceding sections. Several studies are suggestive of relational memory deficits in TLE, although none have directly contrasted relational memory with memory for items. For instance, patients with left TLE did not profit in learning a word list when words belonged to schemas about events, while those with right TLE benefited from these loose a ssociations (Helmstaedter, Gleissner, Di Perna, & Elger, 1997). In a post-surgical sample, pa tients with left TLE performed worse than those with right TLE on memory for word pairs in declarative and nondeclarative formats (Savage, Saling, Davis, & Berkovic, 2002). Similarly, the exte nt of surgical resecti on has been associated with a decreased ability to l earn associations between objects and faces (Weniger, Boucsein, & Irle, 2004). Evaluating relational memory in TLE has the po tential to provide si gnificant information about the role of the hippocampus. Through EEG monitoring and neuroimaging techniques, the complex partial seizures of TLE have been demonstrated to be predominantly of hippocampal origin (Engel, 1996). The most common structural abnormality is unilateral mesial temporal sclerosis (Engel, 1996), which is the only neuropathologic finding in the majority of patients (Trennery, Westerveld, & Meador 1995). Thus TLE provides a na turally occurring model of

PAGE 20

20 hippocampal pathology that can be used to examine hippocampal contributions to memory. Examining relational memory in TLE has the potential not only to provide support for relational memory theory in a new population, but also to improve clinical care. The following sections describe the role of neuropsychol ogy in the clinical ca re of epilepsy and de tail problems with current clinical memory tests used to assess TLE patients. The key idea is that many of these problems can be rectified through the developmen t of more sensitive and specific tests of hippocampal function. The Role of Neuropsychological Testing in the Surgical Treatment of Epilepsy Despite m edical treatment, approximately 30 percent of the total epilepsy population continues to experience seizures that are considered intractable (Helmstaedter, 2004). The seizures associated with mesial TLE are among the most resistant to anti epileptic drugs (Engel, 2001). For treatment refractory TLE cases, surgery to remove a portion of the anterior temporal lobes is an option, giving patients a 70 to 90 percent chance of becoming seizure free (Engel, 2001). The best surgical candidates show unilate ral hippocampal sclerosis, ictal and interictal abnormalities limited to relatively discrete neural zones, and focal/later alized neuropsychological deficits (Helmstaedter, 2004; Loring, 1997). Th e current standard of care for pre-surgical candidates includes neuropsychological testing desi gned to yield information about localization and lateralization of seizure focu s and to help predict whether iatrogenic postoperative cognitive disability will result from the planned resecti on (Helmstaedter, 2004). The degree of deviation from expected pattern of performance on neur opsychological testing may serve as a prognostic indicator of post-operative cognitive decline. While recovery of cognitive function is often noted after surgery, there is also a risk for cognitive decline. Estimates ar e that approximately 40 percent (range 10-60 percen t) of patients experience verbal memory loss after anterior temporal lobectomy (ATL; Chelune, 1995; Loring,

PAGE 21

21 Barr, Hamberger, & Helmstaedter, 2007; Stroup et al., 2003). The risk for cognitive decline is inversely related to the functional adequacy of the tissue to be rese cted in the surgical temporal lobe (Chelune, 1995). For example, post-surgical pa tients tend to show a greater loss in verbal memory if they had better baseline verbal memo ry functioning, which is indicative of more functionally adequate underlying tissue (Helmstaedter, 2004; Str oup et al., 2003). Therefore, neuropsychological testing provide s valuable information that can help physicians and patients inform their decisions about likely outcomes of treatment. Assessment of Memory Functioning in TLE Impaired declarative m emory is a hallmar k symptom of TLE. The material-specific framework guides our understand ing of memory in TLE. Mate rial-specific predictions link memory for verbal material with the language domi nant temporal lobe (usu ally left) and memory for nonverbal material with the language non-dominant temporal lobe (usually right; Milner, 1975). Factor analytic results s uggest that the constr ucts of verbal and nonverbal memory are robust within the TLE population (Davis, Andresen, Witgert, & Breier, 2006). The literature consistently shows a strong link between impaired verbal memory and left TLE (Helmstaedter, 2004; Hermann, Seidenberg Schoenfeld, & Davies, 1997; Strauss et al., 1995) and left hippocampal atrophy (Griffith et al ., 2003; Martin et al., 1999; Trennery, 1996). In contrast, findings are mixed regarding the association between nonverbal memory and right TLE. Research generally has not demonstrated a reliable relationship between performance on nonverbal/visual memory measures and right TLE or right hippocampal volume (Griffith et al., 2003; Helmstaedter, 2004; Mar tin et al., 1999; OBrien, Bowd en, Bardenhagen, & Cook, 2003; Trennery, 1996). While verbal memory is typically assessed by story-recall or list-learn ing paradigms, there is little agreement on wh ich tests best identify n onverbal memory impairments. Currently most

PAGE 22

22 clinical assessments incorporate measures of figural reproduction, such as the Visual Reproductions subtest of the Wechsler Memory Scale-Third Edition (Wechsler, 1997b) or the Rey Complex Figure Test (Meyers & Meyers, 1995) which ask the examinee to draw the figure from memory after a single presentation. A fe w studies have demonstrated some lateralizing value of figural reproduction tests in TLE (Glosser, Cole, Khat ri, DellaPietra, & Kaplan, 2002; Jones-Gotman, 1991). However, the sensitivity of these tests to detect memory dysfunction specific to right TLE is questionable. For instance, a meta-analysi s was conducted on 33 studies of TLE that used the Wechsler Memory Scale subtests of Logical Memo ry and Visual Reproductions to assess verbal and nonverbal memory respectively (Lee, Yip, and Jones-Gotman, 2002). While the verbal task was sensitive to left hemisphere dysfunction both preand post-operatively, the efficacy of the nonverbal task to assess right hemis phere dysfunction was not confirmed. Further compelling evidence of the poor lateralizing capability of figural reproduction tests comes from a study by Barr and colleagues (1997). In a sample of approximately 750 patients, they detected no significant differences betw een patients with right and left TLE on either Wechsler Memory Scale Visual Reproductions or the Rey Complex Figure Test. Their analyses were well powered and controlled for poten tial confounds, limiting any concerns about inadequacies in design. Similar negative findings have been noted in other studies of TLE using the Brief Visuospatial Memory Te st-Revised (Benedict, 1997), a figural reproduction test with multiple learning trials, and th e Continuous Visual Memory Te st (Trahan & Larrabee, 1988), a figural learning and recognition test (Barr, Morrison, Zaroff, & Devinsky, 2004; Snitz, Roman, & Beniak, 1996).

PAGE 23

23 One reason for the poor performance of conventional nonverbal memory tests in this context is that they focus primarily on object/i tem memory, a function th at may not specifically reflect the hippocampal contribution to memory. Other possible reasons exist (see below). Nonverbal tests that tap aspects of spatial memory may be more promising. For instance, several researchers have proposed alternative scoring methods for the Rey Complex Figure. Loring, Lee, and Meador (1988) showed that qualitative scoring cr iteria assessing distortion and misplacement errors discriminated between right a nd left TLE, differences that were not evident when traditional scoring criteria were used. These results were re plicated by Frank and Landeira-Fernandez (2008). Like wise, Breier and colleagues (1996) found that a spatial scoring system for the Rey Complex Figure was more se nsitive to right hippocampal dysfunction than a figural scoring system; however, subsequent studies have not detected differential sensitivity to right TLE using this scoring (Kneebone, Lee, Wade, & Loring, 2007; McConley et al., 2008). Experimental spatial memory tests are also promising. Patients with right TLE have shown worse performance than those with left TLE on a ta sk requiring them to re member spatial details in complex scenes (Baxendale, Thompson, & Pae sschen, 1998) and on a task requiring them to change position while remembering the location of hidden objects in a spatial array (Abrahams et al., 1999). Improving Detection of Memor y Dysfunction in Right TLE As indicated above, there are se veral possible explanations for the failure to demonstrate a consistent link between nonverbal m emory perf ormance and right TLE. Some researchers suggest that nonverbal memory may not be as strictly lateralized or localized as verbal abilities, which could help to account for the findings (Helmstaedter & Kurthen, 2001; Loring et al., 2007). Other explanations focus on test characteri stics that may contribute to their insensitivity to nonverbal memory dysfunction.

PAGE 24

24 The nature of the test stimuli, in particular their level of abstractness, should be considered. Many authors have raised concerns about the verba lizability of stimuli, su ch that nonverbal tests may be contaminated by verbal encoding (Barr et al., 1997). Griffith and colleagues (2003) found that the left hippocampus explained a signi ficant portion of the variance in nonverbal memory performance, which may be considered to support this contention. Similarly, McConley and colleagues (2008) noted modest correlations between left hippocampal volume and figural reproduction performance. Moreover, Helmstaedter, Pohl, and Elger (1995) found that patients with right TLE showed memory impairment for designs only when their complexity exceeded verbal learning capacity. An additional concern is that many measures involve reproduction of designs, which is confounded by motor and constructional abilities. A recognition format would bypass this concern. Another explanation for the mixed findings relates to the type of nonverbal memory assessed. Results described previously indicate that spatial memory tests may be more closely tied to language non-dominant hippocampal function than tests of figural or object memory. These findings could be interpreted in the cont ext of relational memory theory. As spatial memory involves integrating relatio nships to represent the environm ent, it could be considered a subset of relational memory (Cohen et al., 1999). The success of some spatial measures in the identification of neuropsychological morbidity in TLE suggests that other relational memory paradigms may also have clinical assessment value. Most studies on TLE are conducted in clinical settings and use commercially available assessments rather than experimental paradigms. However, commercial tests may not provide the most sensitive measure of hippocampally-mediate d memory function. Measures, such as TI,

PAGE 25

25 that are specifically designed to assess relati onal memory may show better sensitivity to hippocampally-mediated memory function in TLE. Purpose of Current Study The primary aim of the present study was to ex amine TI in patients with intractable TLE who underwent ATL. There are no published studies of TI in TLE, and results have the potential to provide support for relational memory theory in a new population. TI performance and its relationship to side of surgery were examined. It was hypothesized that material-specific deficits in performance would be observed; such that TI for patterned vi sual stimuli would be impaired in patients with right ATL relative to patients with left ATL and healthy controls. Deficits were expected to be most evident on the TI conditi on, although it was expected that right ATL patients may demonstrate deficits on other conditions involving premise pairs from the overlapping hierarchy. In addition, the current study sought to replicate behavioral findings from Heckers and colleagues (2004) study, such as the TI effect for response latencies, which predicts that reaction times will be slowest for pairs involving TI. The secondary aim was to compare the clinical u tility of the TI task to established clinical memory tests. The experimental TI task addr esses some of the con cerns about conventional nonverbal memory tests by utilizing a recognition format and stimuli that are difficult to verbalize. The Rey Complex Figure Test was se lected as the standard measure of nonverbal memory because it is widely used in epilepsy surgery centers to assess memory dysfunction associated with right TLE (Barr et al., 1997; Frank & Landeira-Fernandez, 2008). It was hypothesized that the TI task would provide more accurate detection of right-sided resection than the standard test of nonverbal memory.

PAGE 26

26 CHAPTER 2 METHODS Participants Participants included p atients with intracta ble temporal lobe epilepsy (TLE; n=27), who had undergone a standard anterior temporal lobectomy (ATL), a nd healthy controls (n=24). Participants were 18 years of age or older. Exclusion criteria we re: 1) history of developmental disability or mental retardation resulting in Wech sler Abbreviated Scale of Intelligence (WASI; Psychological Corporation, 1999) FullScale IQ < 70; 2) history of Axis I psychiatric disturbance resulting in hospitalizati on; 3) cranial radiation or chemothera py treatment (within 1 year); or 4) other neurological illness (e.g., ce rebrovascular disease, brain tumor, severe traumatic brain injury). Controls were required to be free of any neurological disease. After completion of testing, three epilepsy patients we re excluded from further analyses Two were eliminated due to concerns about comprehension since English was a second language, a nd another was excluded due to current psychiatric symptoms that resulted in failure to complete the testing session. Procedures Recruitment Strategy The study was approved by the University of Florida Institutional Review Board (#2402007). Epilepsy patients were recruited from the Com prehensive Epilepsy Program at the University of Florida. Patients who had undergone ATL from January 2000 to January 2008 were pre-screened in medical records to determin e if they met inclusion/ exclusion criteria for the study. Seventy-seven patients (38 left ATL, 39 right ATL) met eligibility criteria and were contacted by mail. A letter was sent that inform ed them of the opportunity to participate in a research study and briefly described the study. Enclosed was an addressed, stamped postcard that could be returned, should they desire additional information or want to volunteer.

PAGE 27

27 Potential participants were ma iled up to three times in order to optimize participation. The overall response rate to the mailing was 56%. Seve n patients were unable to be contacted due to change in address. Participation in the study wa s completed by 39% of patients. Twelve patients responded to the mailing but did not participate for various reasons [distance to travel (n=3), failed to meet inclusion criteria after additional screening (n=4 ), declined (n=3), unable to schedule (n=2)]. Healthy contro ls were recruited by asking ep ilepsy patients whether their family members or friends might be interested in participating and also by community flyer. All potential participants underw ent a 10-minute screening by phone regarding demographic and medical information relevant to study eligibility. Assessment Procedures Eligible can didates were sc heduled for testing and assigned a subject number for identification. Participants were tested at University of Florida, in their home, or at the local Epilepsy Foundation in their town of residence. The assessment was administered by a graduate student in psychology or a trained undergraduate research assist ant. The duration of the testing session was three to four hours. Informed consent procedures were completed. Then, a neuropsychological battery consisting of experimental and traditional measures was administered. Participants were compensated $10 per hour for their time, and those tested at the University of Florida were also offered a $3 parking voucher. Demographic information, including age, gender, ethnicity, hande dness, and years of education, was collected. The epilepsy patients we re also asked to provide medical information and to consent to release information from th eir treatment in the Comprehensive Epilepsy Program. Relevant medical information was then collected from the patients medical record. Medical variables included age at onset of epilepsy, family history of epilepsy, current medications, neuroimaging results, seizure late rality and localization, current and previous

PAGE 28

28 seizure frequency, Wada memory support, cerebral speech dominance, date of surgery, type of surgery, and post-surgical outcome. Measures Transitive Inference Task The transi tive inference (TI) task was pres ented on a laptop computer using stimuluspresentation software (E-prime). It was closely modeled after the paradigm developed by Heckers and colleagues (2004). They demonstr ated selective right hippocampal activation during functional imaging of the TI condition, wh ile other task conditions implicated medial temporal areas, such as the anterior parahippoca mpal gyrus. Their stimuli (Heckers et al., 2004) were adapted for use in the current study. They cons ist of thirteen visually distinct pattern fills created from Corel Draw, which were selected to be of similar leve ls of visual interest (Figure 21). The patterns were randomly assigned to pairs of ellipses or pentagons. The stimuli composed a series of four non-overl apping visual stimulus pairs (which will be represented in text by lower case letters a>b, c>d, e>f, g>h) a nd a series of four overlapping visual stimulus pairs (which will be represente d in text by uppercase le tters A>B, B>C, C>D, D>E). During training, participants learned the winner in each pair. They were then tested on their ability to recollect the correct response for previously seen pairs and to infer the correct response for novel pairings. More detailed in formation on training a nd testing procedures follows and is also summarized in Table 2-1. Training The training was designed to ensure that participa nts would learn the correct response for each pairing and would also be likely to hierar chically encode the overlapping stimulus set (Heckers et al., 2004). Training was conducted first on the nonoverlapping pairs and then on

PAGE 29

29 the overlapping pairs. A correct choice was reinfo rced by a smiley face. Sample instructions are provided. For the non-overlapping pairs, participants were instructed: You are going to see pairs of objects on the screen. Press the red key to choose the left object and the yellow key to choos e the right object. Your job is to learn which object is the winner in each pair. When you pick the winner, a smiley face will appear. You won't see the smiley face if you pick the incorrect object. Initia lly you will have to guess which object is the winner in each pa ir. Once you find out which object is the winner in a pair, remember the answer for the next time you see that particular pair. Prior to training for the overlapping pairs, participants were instructed: In this next part, you will see new objects. Ag ain your job is to pick the correct object, the one that produces the smiley face. In this pa rt, a particular object wi ll not be paired with the same partner each time. Whether an object is correct depends on its partner. As before, each time a specific pair of objects is pres ented, the same one will always be correct. However, if an object is paired with a new partner, it may or may not be correct. Participants were also provide d examples before beginning trai ning for each stimulus set. Training for each stimulus set consisted of 144 tr ials, whereby each pair was presented 36 times. The training procedure was separa ted into three blocks (Table 21). The first and second blocks consisted of 60 trials. The first training bloc k was frontloaded to contain twice as many representations of two of the pa irs, while the second training bl ock was backloaded to contain more representations of the other two pairs. Fo r example, participants viewed 20 instances of AB and BC, and 10 instances of CD and DE in th e first training block for the overlapping set. During the second block, particip ants viewed 20 instances of CD and DE, and 10 instances of AB and BC. According to Heck ers and colleagues (2004) the initial front-lo ading of pairs is necessary for healthy adults to correctly make in ferences during the test. The third training block consisted of 24 trials containing equal numbers of the four stimulus pairs. Throughout the training trials, the presentation of the pairs and the position of the two stimuli within each pair were randomized.

PAGE 30

30 Test During testing, me mory for previously viewed pairs and the ability to infer the correct response for novel pairings were assessed. No re inforcement was provided. Participants were instructed: In this section, you will see pairs from Part 1 and Part 2. Your job is to pick the correct object. You will no longer see the smiley face, ev en if you pick the correct object. Pick the object that you think would be correct based on what you learned in practice. In this part, you will also see objects paired in new ways. When this happens, please make your best guess about which object should be correct. Think about what you learned in practice and about the objects in relation to their partners. Th e objects will be on the screen for a limited time, so try to respond as qui ckly and as accurately as you can. Two blocks of testing were administered, and ea ch block consisted of 80 trials divided into 10 trials of a particular type (Trained N on-overlapping, Inference Non-overlapping, Trained Overlapping, Inference Overlapping). Participants were asked to make inferences about five novel pairings from the non-overlapping stimul us set and five novel pairings from the overlapping stimulus set (Table 2-1). An inference from the non-overlapping set always paired an object that had been a winner with an object that had never been reinforced. Therefore, the problem could be solved by simply remembering the reinforcement history of each object. In contrast, inferences from the overlapping set could pair objects that had not been consistently reinforced during training (i.e., object had been winner and loser depending on its partner). For instance, B>D is a critical pair, in which both objects were reinforced on 50% of training trials. In order to solve this transitive problem, it would be es sential to encode the overlapping set as a hierarchy and to remember how a stimul us relates to other s timuli in the hierarchy. Scoring and interpretation Indices derived from perform ance included accu racy and latency scores. A key score was accuracy (percentage correct) for the Overlapping Inference condition, which reflects the process of TI. In the context of the current study, it wa s expected that patient s who had undergone right

PAGE 31

31 ATL would demonstrate significan t difficulty making inferences about visual stimulus pairs from the overlapping set. Latency scores also provided information on relational memory processing. Based on the findings of Heckers and colleagues (2004), significantly longer latencies were expected on responses for ove rlapping vs. non-overlapping pairs and for responses requiring an inference, with la tencies most pronounced on inferences about overlapping pairs. Post-test assessment Following the computer-based task, it was ass ess ed whether participants were explicitly aware of the hierarchical nature of the stimu li (similar to Titone, Ditman, Holzman, Eichenbaum, & Levy, 2004). Participants were presented with two hand-held stimulus cards representing B and D and asked which would be the winner if they were paired. They were then given five cards representing all the stimuli from the overlapping set and aske d to rank order them according to dominance (i.e., which stimulus was most likely to be a winner). The experimenter provided no further information, allo wing participants to deci de what attribute to use to order the cards. Each stimulus in the arrangement was then as signed a difference score, which was derived by subtracting its correct position (1, 2, 3, 4, 5) fr om the position the partic ipant selected (1, 2, 3, 4, 5) and taking the absolute value of the result. For example, a stimulus in the correct position would be assigned a difference score of zero, whereas a stimulus in the third position that should be in the first position would be assigned a di fference score of two. Difference scores were summed to yield a total hierarch ical awareness score (range 0) Lastly, participants were asked whether they were aware that the stimuli were hierarchically organized and to describe their strategies for solving the task.

PAGE 32

32 Standard Neuropsychological Tests Conventional neuropsychological measures we re used to assess general intellectual functioning, verbal and nonverbal me mory, langu age, attention, and ex ecutive function. The following tests are peer-reviewed, well-validated measures to assess these domains. All tests were administered and scored according to standard ized procedures outlined in the test manuals. Resulting performances were demographically corre cted to remove inherent differences due to age. In addition, the normative da ta provided corrections for differe nt educational levels for the measures indicated by an asterix. 1) Wechsler Abbreviated Scale of Intellig ence (WASI; Psychologica l Corporation, 1990). The Vocabulary subtest and the Matrix Reasoning subtest were administered. In Vocabulary, the examinee is asked to orally define words. In Ma trix Reasoning, the examinee is asked to select a pattern to complete an abstract design. Performance on these tw o subtests can be combined to yield an estimate of gene ral intellectual functioning. 2) Hopkins Verbal Learning Test-Revised (HVLT-R; Brandt & Benedict, 2001). The HVLT-R is a measure of the processes invol ved in learning and remembering verbal information. The examinees ability to learn a 12-word list over three trials is examined. Free recall of the list is as sessed after a 20to 25-minute dela y, followed by a recognition trial. 3) Wechsler Memory Scale-Third Edition sele cted verbal subtests (WMS-III; Wechsler, 1997b). In Logical Memory, partic ipants are read two brief stor ies and asked to retell them immediately and after a 30-minute delay. In Verbal Paired Associates, participants are asked to learn a list of eight abstract word pairs over four learning tria ls. During the immediate and 30minute delayed recall, the examiner reads the first word in each pair, and the examinee is asked to recall the second word. Both subtes ts also include a recognition trial.

PAGE 33

33 4) Rey Complex Figure Test a nd Recognition Trial (RCFT; Me yers & Meyers, 1995). The RCFT assesses visuospatial constructional abilit y and nonverbal memory. The test involves copying a complex geometric design, which the pa rticipant then draws fr om memory after a 3minute delay and after a 30-minute delay. A recognition trial follows the delayed free recall. The drawings are scored based on the accuracy and placement of the elements of the design. 5) Boston Naming Test-II (BNT-II; Goodglass & Kaplan, 2000).* This measure of confrontation naming includes 60 black-and-white ink drawings. Pa rticipants are asked to name each picture, and semantic and phonemic cues can be provided. 6) Digit Span subtest from the Wechsler A dult Intelligence ScaleThird Edition (WAIS-III; Wechsler, 1997a). On the Digit Span subtest, exam inees are read a string of digits and asked to recall them in order. The stri ng increases by one after two trials are completed correctly. Then, using the same procedure, the ability to recall digit strings in reverse order is tested. 7) Trail Making Test (Army Individual Test Battery, 1944; Reitan, 1958).* This test measures visuomotor speed, attention, tracking, and setshifting. In Part A, participants are required to connect numbers from 1 to 25 and instruct ed to do so as quickly as possible. In Part B, participants are asked to connect letters and numbers in order, shifting between sets. 8) Wisconsin Card Sorting Test (WCST; Heaton, 1981).* This measure of mental flexibility and problem solving re quires examinees to sort two decks of 64 cards with colored shapes printed on them. The cards should be sorted based on unstated principles, such as color, form, or number. 9) Edinburgh Handedness Inventor y (Oldfield, 1971). This te n-item questionnaire assesses hand preference for a range of activities. A laterality quotient ranging from (left-hand dominance) to +1 (right-hand dominance) can be calculated.

PAGE 34

34 Table 2-1. Protocol for th e transitive inference task Part 1: Training fo r non-overlapping pairs Stimuli pairs Eight ovals with geometric, patterned fills Combined in pairs, with one stimulus in pair arbitrarily designated as winner (a>b, c>d, e>f, g>h) Parameters left/right position counterbalanced random presentation within each block feedback on every trial such that se lecting winner results in smiley face Blocks Block 1: fr ont-loaded, 60 trials ab and cd appear 20 times each; ef and gh appear 10 times each Block 2: back-loaded, 60 trials ef and gh appear 20 times each; ab and cd appear 10 times each Block 3: balanced, 24 trials ab, cd, ef, gh appear 6 times each Part 2: Training for overlapping pairs Stimuli pairs Five pentagons w ith geometric, patterned fills Combined to form a series of overlapping pairs in a hierarchy (A>B, B>C, C>D, D>E) Parameters left/right position counterbalanced random presentation within each block feedback on every trial such that se lecting winner results in smiley face Blocks Block 1: fr ont-loaded, 60 trials AB and BC appear 20 times each; CD and DE appear 10 times each Block 2: back-loaded, 60 trials CD and DE appear 20 times each ; AB and BC appear 10 times each Block 3: balanced, 24 trials AB, BC, CD, DE appear 6 times each Part 3: Test Stimuli pairs previously viewed non-overlapping and overlapping pairs 5 novel pairings from non-overlappi ng set: a>d, a>f, c>f, c>h, e>h 5 novel pairings from overlapping set: A>C, A>D, B>D, B>E, C>E Parameters left/right position counterbalanced no reinforcement Blocks Block 1 and Block 2: Each block is 80 trials divided into sets of 10 trials of a particular type (Non-Overlapping Trained, Non-Overlapping Inference, Overlapping Trained, Overlapping Inference)

PAGE 35

35 Non-overlapping pairs a b c d e f g h Overlapping pairs A B B C C D D E Figure 2-1. Transitive inference stimuli. > > > > > > > >

PAGE 36

36 CHAPTER 3 RESULTS Data were analyzed using the SPSS 16 statistica l software package. An alpha level of .05 was used throughout the analyses. Demographic and Clinical Characterist ics of Participants Descriptive statistics were generated for the total sample and the groups (right ATL, left ATL, control) (Table 3-1). The groups were contrasted on demographic and clinical variables using one-way analyses of variance (ANOVAs), independent samples t-tests, and Pearson chisquare tests. Participants ranged in age from 20 to 66 years ( M = 41.8, SD = 14.0). Educational level ranged from nine to 20 years ( M = 14.9, SD = 2.8). Gender composition of the sample was evenly balanced, with 50% female. The sample was predominantly Caucasian (83%), with a smaller proportion of African-Ameri cans (13%) and persons of Hispanic origin (4%). With the exception of one patient, all pa rticipants reported primarily right-hand dominance as assessed by the Edinburgh Handedness Inventory. No group di fferences were detected on demographic characteristics including age [ F (2, 18.7) = .54, p = .59], education [F (2, 44) = 2.80, p = .07], gender [ 2(2) = .42, p = .81], or race/ethnicity [ 2(4) = 3.83, p = .43]. The epilepsy groups were well matched on disease characteristics (Table 3-1). Mean age at onset of epilepsy was 20.5 years ( SD = 14.1), and the mean duration of epilepsy was 19.6 years ( SD = 14.0). 41.7% of patients had experienced a mild to moderate traumatic brain injury, which may have been a risk factor in the develo pment of their epilepsy. The time from surgery to study evaluation ranged from eight to 95 months, with a mean of 55 months ( SD = 26.8). Regarding surgery outcome, 66.7% of patients were classified as seizure and aura free after surgery. No group differences were detected on clinical variables, including age at onset of epilepsy [ t (22) = -.77, p = .45], duration of epilepsy [ t (22) = 1.10, p = .28], family history of

PAGE 37

37 epilepsy [ 2(1) = .89, p = .37], history of febrile seizures [ 2(1) = .50, p = .48], pre-surgical seizure frequency [ t (22) = .33, p = .75], seizure classification [ 2(1) = .27, p = .60], pre-surgical MRI results [ 2(2) = 2.68, p = .26], time since surgery [ t (22) = -1.34, p = .19], or post-surgical seizure outcome [ 2(2) = .66, p = .72]. Also, no patient had a hi story of vagal nerve stimulator implantation. Data Preparation Score Derivation Scores for accuracy and response latency were derived for performa nce on the transitive inference (TI) task. Accuracy scores were gene rated by taking the mean percentage correct. Latency scores were derived by first elimina ting all response times greater than 15 seconds, which likely reflected the influence of extraneous f actors. Then, for each individual participant, the median response time for correct responses was derived for each part of the task. This resulted in the elimination of 3,765 data points out of 21,504. Accuracy and latency scores were derived for training on each stimulus set (Block 1: Trials 1-48, Block 2: Trials 49-96, and Block 3: Trials 97-144) and for the test conditions: 1) Non-overlapping Trained, 2) Non-overlapping Inference, 3) Overlapping Trained, and 4) Overlapping Inference. Data Screening The assump tions of univariate normality were checked for each of the dependent variables. This preliminary examination showed that distri butions for TI accuracy scores were mildly to moderately negatively skewed. Therefore, squa re-root transformation was performed to improve normality. In subsequent analyses, results usi ng the square-root transformed accuracy data are reported. The analyses were also conducted usin g the raw TI accuracy data, which yielded the same pattern of results as the square-root transf ormed data. The TI reaction times scores showed some evidence of positive skewness and elevated kurtosis for several variables. An outlier

PAGE 38

38 replacement strategy was applied to improve normality of the reaction time distributions. For each variable, scores greater than 2.5 standard deviations above the mean were replaced with the reaction time value at 2.5 standard deviations above the mean. In subsequent analyses, assumptions of the general linear model were te sted when indicated, and appropriate corrections were applied if the assumptions were not met. Aim 1: Transitive Infe rence Task Performance Performa nce on the TI task was evaluated us ing accuracy and latency scores for the training and test conditions. It was hypothesized that infere nces about the overlapping pairs would be impaired in patients who underwent ri ght ATL relative to patients with left ATL and healthy controls. It was expect ed that right ATL patients may s how deficits on other conditions of the task, but deficits would be most evident on the Overla pping Inference condition. Accuracy: Training Descriptive statistics for accuracy (percenta ge correct) are presented in Table 3-2. Training pe rformance for each stimulus set was ev aluated using a 3 (Group: right ATL, left ATL, control) x 3 (Training: Block 1, Block 2, Block 3) mixed between-within ANOVA. First, training for the non-overlapping pairs was examined (Figure 3-1). For the withinsubjects factor, Mauchly s test was significant ( p < .05), suggesting a viola tion of the sphericity assumption and indicating that the variances in the differences betw een blocks were not equivalent. Thus, Greenhouse-Geisser df corrections are reported. Levenes test also reached significance for Block 3 of Training, indicating that the assumption of equality of error variances across groups was not met for this block. The effect of training was significant, F (1.61, 72.62) = 83.56, p < .001, 2 = .65. Bonferroni adjusted post-hoc tests showed significant improvement across training from Block 1 to Block 2 ( p < .001) and from Block 2 to Block 3 ( p < .001). Neither the main effect of group [F (2, 45) = 2.35, p = .11, 2 = .09], nor the interaction between

PAGE 39

39 group and training block [ F (3.23, 72.62) = .56, p = .66, 2 = .02] was significant. Given that the power to detect group differences was only .45, exploratory one-wa y ANOVAs were conducted to further examine performance on each block. Group differences were de tected only on Block 1 of the task, F (2, 45) = 4.34, p = .02, 2 = .16. Patients with right ATL were found to perform more poorly than controls on Block 1, t (45) = 2.73, p = .03. Training performance on overlapping pairs was also evaluated (Figure 3-2). The main effect of training was significant, F (2, 90) = 19.97, p < .001, 2 = .31. Bonferroni adjusted posthoc tests showed that participants improved across training from Block 1 to Block 2 ( p = .001) and from Block 1 to Block 3 ( p < .001). A trend for improvement from Block 2 to Block 3 ( p = .08) was observed. Neither the main effect of group [ F (2, 45) = 2.17, p = .12, 2 = .09], nor the interaction between group and training block [ F (4, 90) = 1.97, p = .10, 2 = .08] was significant. Given that the power to dete ct differences between groups was only .42, exploratory one-way ANOVAs were conducted to further examine pe rformance on each block. Group differences were detected only on Block 2 of the task, F (2, 45) = 3.90, p = .03, 2 = .15. Patients with left ATL performed more poorly than controls on Block 2, t (45) = 2.53, p = .04. Taken together, these findings demonstrate that participants improved across training for both the non-overlapping and overlapping sets. Each group achieved mean accuracy of at least 85% on the third training block for non-overlapping pairs and at least 78% on the third training block for overlapping pairs. Gr oups achieved fairly comparable levels of learning for the nonoverlapping and the overlapping stimulus sets by the third training block. Accuracy: Test Test performa nce was evaluated using a 3 (G roup: right ATL, left ATL, control) x 4 (Condition: Non-overlapping Trained, Non-overlapping Inference, Overlapping Trained, Overlapping Inference) mixed between-within ANOVA. For the within-subjects factor,

PAGE 40

40 Mauchlys test was significant ( p < .05), suggesting a violation of the sphericity assumption. Thus, Greenhouse-Geisser df corrections are repo rted. A significant effect of condition was observed, F (1.76, 79.06) = 6.83, p = .003, 2 = .13. Bonferroni adjusted post-hoc tests revealed greater accuracy on the Non-overlapping Trained condition than on either Overlapping Trained ( p = .01) or Overlapping Novel ( p = .02) conditions (Table 3-2). The main effect of group was also significant, F (2, 45) = 3.94, p = .03, 2 = .15. Bonferroni adjusted post-hoc analyses showed that patients with right ATL pe rformed more poorly than controls (p = .03) on the task (Figure 3-3). The interaction between group and condition was not significant, F (3.51, 79.06) = .73, p = .56, 2 = .03. The primary hypothesis was that patients with right ATL would show selective difficulty with the TI condition (i.e., infe rences about overlapping pairs). Therefore, given the limited power (.21) of the preceding analysis to detect an interaction, expl oratory one-way ANOVAs were conducted to further evaluate performance on the test conditions. Group (right ATL, left ATL, control) was the between-subjects factor an d accuracy score for the test condition was the dependent variable. Group di fferences were not observed on Non-overlapping Trained [ F (2, 45) = 1.71, p = .19, 2 = .07] or Non-overlapping Novel [ F (2, 45) = 1.93, p = .16, 2 = .08] conditions. A trend toward group differences was detected on the Overlapping Trained condition [ F (2, 45) = 2.68, p = .08, 2 = .11], with right ATL performing mo re poorly than controls. The only test condition in which significant group di fferences were detected was the Overlapping Inference condition [ F (2, 45) = 4.96, p = .01, 2 = .18] (Figure 3-3). Bonferroni adjusted posthoc tests showed that patients with right ATL performed worse than controls, t (45) = 3.15, p = .009. This finding supports the hypothesis that patients would have selective difficulty with nonverbal TI after right ATL.

PAGE 41

41 Level of initial learning was considered as a source of variability in test performance. Although groups achieved fairly comparable levels of learning during trai ning, differences were observed on several individual trai ning blocks. Therefore, analys es of covariance (ANCOVAs) to control for initial learning were conducted. The covariate wa s accuracy on the third training block for the appropriate stimulus set. Ag ain, group was the between-subjects factor and accuracy score for the test condition was the depe ndent variable. These analyses revealed the same pattern as reported in th e previous analysis. For NonOverlapping Inference, initial learning was significantly re lated to performance, F (1, 44) = 84.16, p < .001, 2 = .66. There was no main effect of group, F (2, 44) = 1.10, p = .34, 2 = .05. For Overlapping Inference, initial learning was also significantly related to performance, F (1, 44) = 45.43, p < .001, 2 = .51. A main effect of group was observed, F (2, 44) = 4.93, p = .01, 2 = .18. Bonferroni adjusted post-hoc tests revealed that patients with right ATL performed worse than controls, t (44) = 2.99, p = .01, on this TI condition. Intra-group variability in test performance wa s observed, as evidenced by large standard deviations (Table 3-2). Demogr aphic and clinical variables were examined to determine whether any were related to test performance (Table 33). Age was significantly related to performance in three of the four test conditions, such that older age was associated with lower accuracy scores. The relationship between ag e and test performance did not appear to be an artifact of an underlying relationship between dura tion of epilepsy and test performance as these two variables were not significantly correlate d. Moreover, within the epileps y sample, a series of partial correlations controlling for durati on of epilepsy revealed moderate relationships between age and these test conditions (r = .39 to .50).

PAGE 42

42 Therefore, ANCOVAs were run for these three test conditions, with group as the betweensubjects factor, accuracy score as the dependent variable, and age as the covariate. Age was a significant covariate in all the analyses ( p < .05). Only on the Over lapping Inference condition was a main effect of patient group observed, F (2, 44) = 4.54, p = .02, 2 = .20. Again, Bonferroni adjusted post-hoc tests showed that patients with right ATL performed worse than controls, t (44) = 3.00, p = .01, on this TI condition. In addition, seizure freedom wa s considered as a potential influence on test performance (Table 3-3). Two left ATL patie nts and two right ATL patients continued to have seizures after surgery, which is a less successful outcome and indicates that epileptogenic tissue remains in their brains and is potentially disturbing cognitive f unctions. Test performance was re-examined when these four patients were excluded fr om the sample. ANCOVAs with group as the between-subjects factor, accuracy score for the test as the depe ndent variable, and age as the covariate revealed the same patte rn of results observed in the larger sample. These analyses, coupled with the non-significant correlations be tween seizure freedom and test performance, suggest that including these patien ts in the sample should not have a major impact on the results. In light of the poor performance of right ATL patients on the Overlapping Inference condition, an examination of individual inference pairs was made (Figure 3-4). The critical inference pair, BD, contained stimuli that had both been reinforced 50% of the time during training. The remaining four inference pairs cont ained end items and would be expected to have a lesser degree of difficulty. Thus, the largest disparity between groups would be expected on the BD pair. In view of the small sample and skewed distributions of some scores, performance on individual pairs was examined with the Kruskal-Wallis H test, which is the nonparametric analogue to the between-subjects ANOVA. Gr oup was the independent variable, and accuracy

PAGE 43

43 for the pair was the dependent variable. Accura cy for the pairs, AC, AD, and BE, did not differ significantly between groups (p > .35). Group differences we re observed on the BD pair [ 2(2) = 6.31, p = .04, 2 = .13] and the CE pair [ 2(2) = 6.24, p = .04, 2 = .13]. Post-hoc analyses, using either the Mann Whitney U test or rank sums test, show ed that patients who underwent right ATL performed more poor ly than controls on BD ( p < .05). Post-hoc tests revealed no significant group differences on the CE pair. Hierarchical Awareness Following the computer-based task, it was ass ess ed whether participants were aware of the hierarchy in the overlapping stimul us set. First, they were presented with hand-held cards representing B and D stimuli from the overlappi ng set and asked to select the winner. The proportion of participants who correctly selected B was 56.2% of patients with right ATL, 75.0% of patients with left ATL, and 87.5% of controls A chi-square analysis revealed that these differences were not statistically significant, 2(2) = 5.00, p = .08, but showed a trend in the predicted direction. Participants were then presented with five hand-held cards representing the stimuli from the overlapping set and asked to arrange them in order of dominance. A hierarchical awareness score (range 0) was derived, with lower scores indicative of greater hier archical awareness. The hierarchical awareness score was relate d to performance on the Overlapping Inference condition ( r = .56, p < .001), after controlling for the influence of age and initial learning. Awareness of the hierarchy accounted for 32% of the variance in scores on the Overlapping Inference condition. Hierarchical awareness scores were calc ulated for each group, control ( M = 3.1, SD = 2.9), left ATL ( M = 3.7, SD = 4.6), and right ATL ( M = 4.4, SD = 2.9). A one-way ANOVA with group as the between-subjects factor and hierarchical awareness as the dependent variable was performed. No gr oup differences were detected, F (2, 45) = .76, p = .47, 2 = .03.

PAGE 44

44 After the ordering task, particip ants were asked whether they were aware that the stimuli in the overlapping set formed a hierarchy. The proportion of partic ipants who endorsed hierarchical awareness was 31.2% of patients with right ATL, 42.9% of patients with left ATL, and 62.5% of controls. A chi-squa re test indicated that these did not represent significant group differences, 2(2) = 3.87, p = .14. Reaction Time: Training Descriptive statistics for median reaction tim e are presented in Table 3-4. Training performance for each stimulus set was evaluated usi ng a 3 (Group: right ATL, left ATL, control) x 3 (Training: Block 1, Block 2, Bl ock 3) mixed between-within ANOVA. First, training on the non-overlapping stimul us set was evaluated (Figure 3-5). For the within-subjects factor, Mauchl ys test was significant ( p < .05), suggesting a violation of the sphericity assumption. Thus, Gr eenhouse-Geisser df corrections are reported. The effect of training was significant [ F (1.31, 59.11) = 45.50, p < .001, 2 = .50]. Bonferroni adjusted posthoc tests showed improved reaction times from Block 1 to Block 2 ( p < .001) and from Block 2 to Block 3 ( p = .02). There was also a main effect of group [ F (2, 45) = 3.63, p = .03, 2 = .14]; however, Bonferroni-corrected post-hoc analyses did not show significant group differences. Post-hoc tests that were not adjusted for multiple comparisons revealed that controls had faster reaction times than patients with left ATL ( p = .02) or right ATL ( p = .05). The interaction between group and training bl ock was not significant [ F (2.63, 59.11) = .23, p = .85, 2 = .01]. Given that the power to dete ct differences between groups was only .64, exploratory one-way ANOVAs were conducted to further examine pe rformance on each block. Group differences were detected on Block 3 [ F (2, 45) = 6.08, p = .005, 2 = .21]. Control subj ects demonstrated faster reaction times on Block 3 than patients with left ATL, t (45) = 3.01, p = .01, and patients with right ATL, t (45) = 2.62, p = .04.

PAGE 45

45 Training performance on overlapping pairs was also evaluated (Figure 3-6). The main effect of training was significant [ F (2, 90) = 25.74, p < .001, 2 = .36]. Bonferroni adjusted posthoc tests showed that reaction times improve d across training from Block 1 to Block 2 ( p < .001) and from Block 1 to Block 3 ( p < .001). Neither the main effect of group [ F (2, 45) = 3.03, p = .06, 2 = .12], nor the interaction be tween group and training block [ F (4, 90) = 1.14, p = .34, 2 = .05] was significant. Given the trend toward group differences showing right ATL performing worse than controls and the limited power (.56) to detect group differences, exploratory one-way ANOVAs were conducted to examine reaction ti mes for each block. A trend toward group differences was observed on Block 1 [ F (2, 45) = 2.92, p = .06, 2 = .11] and Block 3 [ F (2, 45) = 2.91, p = .06, 2 = .11]. In summary, participants demonstrated quick er reaction times as training progressed for both the non-overlapping and the overlapping sets. On the final training block for nonoverlapping pairs, controls showed faster reaction times than eith er of the two epilepsy groups. No significant group differences were observed on reaction times during training for the overlapping pairs. Reaction Time: Test Test performa nce was evaluated using a 3 (G roup: right ATL, left ATL, control) x 4 (Condition: Non-overlapping Trained, Non-overlapping Inference, Overlapping Trained, Overlapping Inference) mixed between-within ANOVA. The main effect of group was not significant, F (2, 43) = 2.01, p = .15, 2 = .08, (Figure 3-7). There was a main effect of condition, F (3, 129) = 6.90, p < .001, 2 = .14. Bonferroni adjusted post-hoc tests showed that participants exhibited significantly slower reaction times on the Overlapping Trained condition compared to the Non-overlapping Trained (p = .001) and Non-overlapping Inference (p = .006) conditions. The main effect was moderated by a significant interaction between group and condition, F (6,

PAGE 46

46 129) = 2.86, p = .01, 2 = .12. Decomposing this interaction re vealed that control participants had faster reaction times on the Non-overlapping Trained condition co mpared to either patients with right ATL ( p = .04) or left ATL ( p = .04). Participants exhibi ted comparable reaction times on conditions involving the overlapping stimulus set. In addition, reaction times on the te st were analyzed to ascertain if the TI effect reported by Heckers and colleagues (2004) could be replicated. The TI effect would be demonstrated by an interaction between sequence and inference, su ch that reaction times would be slowest on responses requiring inference across the overlappi ng pairs. Reaction times on the test were analyzed using 2 (Sequence Type: overlapping, non-overlapping) x 2 (Inference Type: present, absent) repeated measures ANOVAs. In the control group, a main effect of sequence was observed, F (1, 22) = 28.61, p < .001, 2 = .56, such that controls s howed significantly faster reaction times on the non-overlapping compared to the overlapping pairs. Neither the main effect of inference [ F (1, 22) = 1.01, p = .33, 2 = .04] nor the interaction between sequence and inference was significant [ F (1, 22) = .59, p = .45, 2 = .03]. This analysis was also conducted in the epilepsy sample. In the epilepsy patie nts, a main effect of sequence was observed, F (1, 22) = 6.13, p < .02, 2 = .22, such that patients also showed si gnificantly faster reaction times on the non-overlapping compared to the over lapping pairs. Again neither the main effect of inference [ F (1, 22) = .16, p = .69, 2 = .01] nor the interaction betw een sequence and inference was significant [ F (1, 22) = .13, p = .72, 2 = .01]. Aim 2: Comparison of Transitive Inference and Standard Neuropsychologica l Tests The secondary aim of the study was to compare the clinical utility of the TI task to established clinical memory tests. It was hypot hesized that the TI ta sk would provide more sensitive detection of language non-domina nt side of surgery than standardized neuropsychological tests of memory. Results from the previous section showed that accuracy on

PAGE 47

47 the Overlapping Inference test condition appeared mo st sensitive to right A TL. Therefore, this TI score was selected to represent performance in the subse quent analyses for this aim. Standard Neuropsychological Test Performance Performa nce on standard neuropsychological test s was examined using a series of one-way ANOVAs with group as the independen t variable (right ATL, left ATL, control) and test score as the dependent variable (Table 3-5). Groups were comparable on intellectual functioning, which was in the average range. As expected, significant group differences were observed on verbal memory tests, including scores for HVLT-R Total [ F (2, 44) = 7.63, p = .001, 2 = .26], HVLT-R Delayed Recall [ F (2, 44) = 6.51, p = .003, 2 = .23], WMS-III Logical Memory Total [ F (2, 43) = 4.39, p = .02, 2 = .17], WMS-III Verbal Pa ired Associates Total [ F (2, 44) = 9.53, p < .001, 2 = .30], and Verbal Paired Associates Delayed Recall [ F (2, 42) = 5.49, p = .008, 2 = .21]. Group differences were also observed on the Boston Naming Test-II [ F (2, 41) = 7.51, p = .002, 2 = .27]. Bonferroni-corrected post-hoc tests showed that patients with left ATL performed worse than controls on these verbal memory scores and on the Boston Naming Test-II ( p < .05). Furthermore, the HVLT-R (Total and Delaye d Recall) and the WMS-III Verbal Paired Associates (Total) discriminated between side of surgery, such that patients with left ATL performed significantly worse than those with right ATL ( p < .05). In contrast, group differences were not observed on a measure of nonverbal memory, the Rey Complex Figure Test. Mean scores for im mediate and delayed recall on this measure ranged from the 5th to the 20th percentile acr oss groups. Also, group differences were not detected on measures of atte ntion or executive function.

PAGE 48

48 Relationship between Transitive Inferen ce and Standard Neuropsychologica l Tests A series of partial correlati ons controlling for age examined relationships between the experimental task and neuropsyc hological test performance (Tab le 3-6). Both non-TI (i.e., inferences about non-overlapping pairs) and TI (i.e., inferences about overlapping pairs) were examined. All significant corr elations were positive and rang ed from small to moderate ( r = .3 to .4). Non-TI was significantly correlated with two verbal memory measures (WMS-III Logical Memory and Verbal Paired Associates) and severa l measures of attenti on/ executive functioning (WAIS-III Digit Span, Trail Making Test Part B, and Wisconsin Card Sorting Test Errors). In addition, non-TI and TI were correlated with the copy condi tion of the Rey Complex Figure Test, a visuospatial, constructional measure. TI was significantly correlated with WAIS-III FullScale IQ score, WMS-III Logical Memory (Total and Delayed), the Rey Complex Figure Test (Immediate and Delayed), and Trail Making Test Part B. Prediction of Right vs. Left Side of Resection Binary logistic regression wa s conducted to evaluate neuropsyc hological predictors of side of surgery. Two neuropsychological scores were entered as predictors, the Rey Comp lex Figure Test Delayed Recall raw score and th e TI score. Age was also specifi ed as an a priori predictor. The model achieved an overall classification accuracy of 73.9%, which is not a significant improvement [ 2(3, 23) = 3.03, p = .39] over the base rate classification of 65%. While the model correctly classified 93.3% of those who unde rwent right ATL, it correctly classified only 37.5% of those who underwent left ATL. Odds ratios (i.e., post-test probability of having undergone right ATL) were calculated as .97 fo r the Rey Complex Figure score and 1.43 for the TI score. Receiver operating characteristic (ROC) curve analyses were performed to determine cut scores that minimize diagnostic errors. The ROC cu rve analyses plot the rate of false positives

PAGE 49

49 (i.e., false prediction of right ATL in someone with left ATL) against the rate of true positives (i.e., correct classification of right ATL). The area under th e curve represents how well each measure predicts right-sided re section, with larger areas indicative of stronger predictors. ROC curve analysis for the Overlapping Inference score revealed area under the curve of .68 (SE = .12, 95% CI = .44.91) (Figure 3-8). A cut score of 75% yielded the following operating characteristics: sensitivity = 50%, specificity = 87.5%, positive predictive power = 89%, and negative predictive power = 47%. Thus, while a score below the cut-off indicated a high probability that a patient had undergone ri ght ATL, only half the patients who underwent right ATL were detected with this cut score. Using a higher cut score identified more patients who underwent right temporal lobe surgery, but yielded an unacceptably high rate of false positives. For example, a cut-score of 85% yi elded sensitivity = 62.5%, specificity = 62.5%, positive predictive power = 76.9%, and negative predictive power = 45.5%. To further examine TI performance, independent samples t-tests and Pearson chi-square tests contrasted right ATL patients who performed above (n=8) and below (n=8) the 75% cut score. No significant group differences ( p > .10) were detected on demographic or clinical variables, including age, years of education, age at onset of epilepsy, duration of epilepsy, months since surgery, language or memory domina nce, or presence of pre-surgical hippocampal sclerosis. Group differences were detected on self-reported hier archical awareness [ 2(1) = 7.27, p = .007], such that 100% of patients who perfor med below the cut score denied hierarchical awareness, while only 37.5% of patients who performed above the cut score did not endorse it. ROC curve analysis was also conducted to ev aluate the predictive utility of the Rey Complex Figure Test (Fi gure 3-8). This analysis used the age-corrected score for Delayed Recall. Area under the curve was .66 (SE = .11, 95% CI = .43.88). Operating characteristics

PAGE 50

50 were calculated for cut scores at T = 30 (s ensitivity = 40%, specifi city = 100%, positive predictive power = 100%, negative predictive po wer = 47%) and at T = 35 (sensitivity = 60%, specificity = 50%, positive predictive power = 69%, negative predictive power = 40%).

PAGE 51

51 Table 3-1. Demographic and clinical characteristics by group Right ATL (n=16) Left ATL (n=8) Control (n=24) Age 43.9 (11.1) 43.9 (15.5) 39.7 (15.4) Years of educationa 14.3 (2.7) 13.4 (2.1) 15.7 (2.9) Gender (% female) 44% 50% 54% Race (% Caucasian) 87.5% 100% 75% History of mild-moderate TBI 50.0% 25.0% 0% History of psychiatric diagnosis 25.0% 25.0% 0% Handedness (laterality q uotient) 0.84 (.18) 0.52 (.67) 0.86 (.16) Age at onset of epilepsy (yrs.) 22.1 (14.4) 17.4 (13.8) Duration of epilepsy (yrs.) 17.4 (12.3) 24.0 (16.9) Family history of epilepsy 31.2% 50.0% History of febrile seizures 25.0% 12.5% Pre-surg. seizure frequency (per month) 8.4 (11.5) 10.1 (11.4) Seizure classification Complex partial Generalized 93.7% 6.2% 87.5% 12.5% Pre-surgical MRI results No abnormalities Hippocampal sclerosis Other lesion/ abnormality 37.5% 56.2% 6.2% 12.5% 62.5% 25.0% Language dominance (Wada) b Left Mixed 100% 0% 71% 29% Months since surgery 60.1 (27.7) 44.9 (23.2) Post-surgical seizure freedom Seizure free Aura only Continued seizures 68.7% 18.7% 12.5% 62.5% 12.5% 25.0% Note. Data presented as mean (standard deviation) except for those variable s listed as percent of sample. aIn control group, one subject was missing years of education. bOne left ATL and two right ATL patients missing Wada language dominance data.

PAGE 52

52 Table 3-2. Transitive inference accuracy scores as percent correct by group Right ATL (n=16) Left ATL (n=8) Control (n=24) Training: Non-overlapping pair s 77.9 (20.2) 81.4 (10.5) 88.1 (15.6) Block 1 66.0 (16.6) 68.7 (12.7) 78.9 (16.3) Block 2 82.2 (23.5) 83.9 (14.3) 91.2 (18.4) Block 3 85.4 (24.4) 91.7 (8.6) 94.1 (15.0) Training: Overlapping pair s 74.6 (15.6) 73.0 (10.5) 81.7 (12.9) Block 1 69.5 (13.1) 69.8 (8.0) 73.0 (14.9) Block 2 76.3 (18.4) 71.1 (12.3) 85.8 (15.2) Block 3 77.9 (18.2) 78.1 (17.2) 86.4 (13.3) Test Non-overlapping learned 85.0 (22.5) 90.6 (12.6) 94.1 (15.7) Non-overlapping inference 82.0 (28.4) 82.8 (23.6) 92.8 (20.6) Overlapping learned 76.2 (16.9) 82.2 (17.5) 87.2 (15.3) Overlapping inference 71.2 (22.1) 82.2 (23.7) 89.1 (15.5) Note. Data presented as mean (standard deviatio n) of percentage correct, with 50% representing chance performance and 100% representing error-less performance.

PAGE 53

53 50 55 60 65 70 75 80 85 90 95 100 1 to 4849 to 9697 to 144Training blockPercent correct RATL LATL Control Figure 3-1. Accuracy for non-overl apping pairs by training block. 50 55 60 65 70 75 80 85 90 95 100 1 to 48 49 to 96 97 to 144Training blockPercent correct RATL LATL Control Figure 3-2. Accuracy for overla pping pairs by training block.

PAGE 54

54 50 55 60 65 70 75 80 85 90 95 100 Non-overlapping trained Non-overlapping inference Overlapping trained Overlapping inferenceTest conditionPercent correct RATL LATL Control Figure 3-3. Accuracy by test condition. 50 55 60 65 70 75 80 85 90 95 100 ACADBDBECETransitive inference pairsPercent correct RATL LATL Control Figure 3-4. Accuracy for overlapping pairs requiring transitive inference.

PAGE 55

55 Table 3-3. Correlations between TI test, demographic characteristics, and clinical variables Nonoverlapping trained Nonoverlapping inference Overlapping trained Overlapping inference Age -.25 -.33* -.47** -.46** Years of education .18 .25 .17 .20 Age at onset (yrs.) -.30 -.38 -.10 -.13 Duration of epilepsy (yrs.) -.06 -.13 -.26 -.18 Seizure freedom .26 .34 .02 .22 Note. Square-root transformed accuracy scores were used in the Pearson bivariate correlations. *p < .05. **p < .01. Table 3-4. Transitive infe rence reaction times by group Right ATL (n=16) Left ATL (n=8) Control (n=24) Training: Non-overlapping pairs 1177.8 (473.9) 1339.9 (294.9) 908.7 (295.7) Block 1 1470.5 (587.7) 1592.4 (321.2) 1260.1 (506.6) Block 2 1143.2 (474.2) 1272.9 (257.9) 905.4 (360.9) Block 3 1075.1 (445.4) 1205.7 (363.7) 788.5 (235.5) Training: Overlapping pair s 1436.8 (476.5) 1377.6 (286.5) 1147.2 (364.7) Block 1 1644.8 (646.0) 1706.7 (349.6) 1304.9 (469.0) Block 2 1411.4 (512.6) 1310.6 (302.3) 1133.0 (369.7) Block 3 1366.7 (488.9) 1223.7 (343.4) 1058.7 (344.6) Test Non-overlapping learned 1163.5 (336.6) 1243.4 (355.7) 865.4 (323.6) Non-overlapping inference 1217.0 (463.7) 1262.4 (575.6) 871.1 (544.6) Overlapping learned 1343.4 (510.7) 1261.4 (363.7) 1210.7 (524.0) Overlapping inference 1404.3 (577.8) 1148.9 (341.2) 1156.0 (524.3) Note. Data presented as mean (standard deviatio n) of median response time in milliseconds for correct responses.

PAGE 56

56 500 700 900 1100 1300 1500 1700 1900 1 to 48 49 to 96 97 to 144Training blockReaction time (msec) RATL LATL Control Figure 3-5. Reaction time for non-overlapping pairs by training block. 500 700 900 1100 1300 1500 1700 1900 1 to 4849 to 9697 to 144Training blockReaction time (msec) RATL LATL Control Figure 3-6. Reaction time for ove rlapping pairs by training block.

PAGE 57

57 500 600 700 800 900 1000 1100 1200 1300 1400 1500 Non-overlapping trained Non-overlapping inference Overlapping trained Overlapping inferenceTest conditionReaction time (msec) RATL LATL Control Figure 3-7. Reaction time by test condition. A B Figure 3-8. Receiver operating ch aracteristic (ROC) curves for th e prediction of right anterior temporal lobectomy. A) Transitive inference (acc uracy for overlapping pairs). B) Rey Complex Figure Test delayed recall.

PAGE 58

58 Table 3-5. Neuropsychologica l test performance by group Right ATL (n=16)a Left ATL (n=8)b Control (n=24)c F p WASI Full-Scale IQ (SS) 101.5 (13.0) 102.1 (9.9) 104.4 (10.8) .30 .74 HVLT-R Immediate total (T) Delayed recall (T) 41.7 (10.2) 42.4 (13.6) 30.0 (9.4) 26.1 (8.0) 44.3 (7.8) 43.2 (11.8) 7.63 6.51 .001 .003 WMS-III Logical Memory Immediate total (ss) Delayed recall (ss) 9.4 (2.1) 10.1 (2.6) 8.4 (3.4) 9.7 (2.6) 11.3 (2.8) 11.6 (2.4) 4.39 2.50 .02 .09 WMS-III Verbal Paired As. Immediate total (ss) Delayed recall (ss) 10.5 (2.1) 10.9 (2.6) 6.9 (3.1) 8.3 (3.8) 11.1 (2.3) 11.8 (1.9) 9.53 5.49 <.001 .008 Rey Complex Figure Test Copy (raw) Immediate recall (T) Delayed recall (T) Recognition (T) 29.7 (4.8) 34.9 (13.6) 34.1 (12.7) 41.7 (10.1) 31.8 (2.2) 41.7 (9.2) 39.9 (9.3) 43.7 (5.3) 31.0 (4.2) 38.3 (15.0) 37.6 (12.0) 48.4 (11.5) .81 .67 .70 2.03 .45 .51 .50 .14 Boston Naming Test-II (T) 43.2 (9.8) 34.1 (4.1) 51.3 (13.3) 7.51 .002 WAIS-III Digit Span (ss) 10.9 (3.3) 9.9 (4.0) 11.6 (3.2) .75 .48 Trail Making Test Part A (T) Part B (T) 39.9 (12.3) 47.1 (8.5) 46.4 (11.2) 50.0 (9.3) 43.9 (10.4) 47.9 (12.3) 1.03 .19 .36 .83 Wisconsin Card Sorting Test Errors (T) Perseverations (T) 45.2 (10.9) 45.1 (10.3) 50.4 (10.6) 53.3 (12.7) 50.6 (7.6) 51.4 (8.7) 1.38 2.04 .26 .14 Note. SS=Standard Score, T=T score, ss=scaled score. aMissing data for right ATL included two Full-Scale IQ, one Digit Span, four WCST, one Trail Making Test, three Bost on Naming Test, one Verbal Paired Associates Delayed, one Logical Memory, and one Rey Complex Figure Test. bMissing data for left ATL included one Full-Scale IQ, one WCST, and one Verbal Paired Associates Delayed. cOne control was only administered experimental measures. Two controls were missing the Full-Scale IQ, and two controls were not administered the WCST.

PAGE 59

59 Table 3-6. Partial correlations between transitive and non-transitive inference and performance on standard neuropsychological te sts after controlling for age Non-transitive inference Transitive inference r r WASI Full-Scale IQ .26 .41** HVLT-R Immediate total Delayed recall -.10 .11 .21 .21 WMS-III Logical Memory Immediate total Delayed recall .31* .35* .32* .39** WMS-III Verbal Paired Associates Immediate total Delayed recall .30* .07 .24 .14 Rey Complex Figure Test Copy (raw) Immediate recall Delayed recall Delayed recognition .37* .14 .14 .26 .33* .33* .32* .27 Boston Naming Test-II .26 .28 WAIS-III Digit Span .40** .16 Trail Making Test Part A Part B -.04 .39** .24 .38* Wisconsin Card Sorting Test Errors Perseverations .34* .26 .24 .22 Note. Square-root transformed accuracy scores represent perf ormance on the TI task. Standard, scaled, or T scores represent performance on standard neuropsychological measures unless otherwise noted. N=47. Missing data included five Full-Scale IQ, one Digit Span, seven WCST, one Trail Making Test, three Boston Naming Test, two Verbal Paired Associates Delayed, one Logical Memory, and one Rey Complex Figure Test. p < .05. ** p < .01.

PAGE 60

60 CHAPTER 4 DISCUSSION Transitive Inference Performance in Temporal Lobe Epilepsy This study examined perfor mance on a transitive inference (TI) paradigm in patients who had a history of intractable TLE and who had under gone standard ATL for seizure relief. During training on the task, participants demonstrated learning of the visual stimulus pairs through improved accuracy and faster reaction times as tr aining progressed. Epilepsy patients achieved similar levels of accuracy as controls. On the test, patients who had undergone right ATL performed significantly worse than control part icipants, and patients who had undergone left ATL performed in the intermediate range. Si gnificant performance differences were not observed between patients based on side of surgery. Importantly, the right ATL group performe d more poorly on the TI condition, which involved making inferences across a series of overl apping pairs that formed a hierarchy. On the TI condition, patients with righ t ATL achieved a mean score of only 71% correct, while patients with left ATL and control participants achieved mean scores of 82% and 89% respectively. The effect size for this group difference was in th e medium range, and this difference remained evident after controlling for the effects of age and initial learning. This finding supports the primary hypothesis that patients wi th right ATL would have sele ctive difficulty with TI for patterned shapes. As expected, the true TI pa ir (BD), in which both stimuli had been equally reinforced during training, was the most difficult individual pair for right ATL patients. An examination of demographic and clinical variables showed that age was significantly related to accuracy scores. Older individuals te nded to perform more poorly, which is consistent with prior research indicating that older adults experience difficulty with the relational organization of propositions within memory, al though previous studies had employed samples

PAGE 61

61 with considerably older ages than were represen ted in the current study (R yan, Moses, & Villate, 2009). Other demographic and clinical variables, including level of educ ation, age at onset of epilepsy, duration of epilepsy, and post-surgical seizure freedom, did not show a significant relationship with accuracy scores. On reaction times for the test, the only group difference noted was faster performance by controls than either epilepsy patient group on the Non-overlapping Trained condition. This finding likely reflects reductions in cognitive and psychomotor speed associated with chronic temporal lobe epilepsy (Hermann et al., 2006) and with the use of antiepileptic drugs (Motamedi & Meador, 2004). Moreover, results showed that the non-overlapping conditions may have been easier for participants, as evidenced by highe r accuracy scores (Non-overlapping Trained > Overlapping Trained, Overlapp ing Inference) and faster reaction times (Non-overlapping Trained, Non-overlapping Inferen ce > Overlapping Trained). Since memory demands were less for these conditions, they may have had a greate r likelihood of detecting differences in basic, speeded processing capacities. After the computerized task, hierarchical aw areness was measured by asking participants to order the stimuli in the overlapping set. Their ability to demons trate knowledge of the hierarchy accounted for 32% of the varian ce in performance on th e TI condition, after controlling for the effects of ag e and initial learning. Significa nt group differences in selfreported hierarchical awareness were not obser ved, although twice as many controls (62%) as right ATL patients (31%) endorsed hierarchical awareness. Li bben and Titone (2008) reported that participants who are aware of the hierarchy may be more lik ely to use a logic-based strategy to solve the transitive problem, while participan ts who are unaware may be more likely to use a stimulus-driven strategy.

PAGE 62

62 The current study utilized a TI paradigm closely modeled after the task developed by Heckers and colleagues (2004). The results partially replicated their behavi oral data. In their healthy, young adult sample, par ticipants achieved a mean of greater than 90% on all test conditions. Healthy controls in the current study achieved simila r mean scores, ranging from 87 to 94% on the various test conditions, while the sc ores of epilepsy patient tended to be lower. Regarding reaction time data, Heckers and collea gues (2004) noted a main effect of sequence (i.e., reaction times faster for non-overlapping than for overlappi ng pairs), a main effect of inference (i.e., reaction times quicker for responses that did not require an inference), and an interaction between sequence and inference (i .e., the TI effect: reaction times slowest on responses requiring an inference across overlapping pairs). In th e current study, the effect of sequence was replicated, but no main effect of in ference nor interaction effect was observed in either control participants or epilepsy patients. Transitive Inference and Cognitive Neuroscience Research The post-surgical ATL population provides a model to study hippocam pal contributions to memory, and the current results suggest a link between hippocampal function and TI. Functional imaging studies have identified a distributed neural network invol ved in TI judgments, including areas of the cortex, hippocampus, and thalamus (Heckers et al., 2004; Zalesak & Heckers, 2009). This research has consistently highlighted the integral role of the hippocampus in TI and has demonstrated selective hippocampal activation duri ng TI of geometric designs (Heckers et al., 2004; Zalesak & Heckers, 2009), faces (Nagode & Pardo, 2002), and face-house pairings (Preston et al., 2004). The transla tional paradigm of TI is also supported in studies of animals, which showed that hippocampally-lesioned rodents exhibit impaired TI (Devito et al., in press; Dusek & Eichenbaum, 1997).

PAGE 63

63 Functional imaging studies have indicated that encoding and recall of associative memories are functions of an integrated hi ppocampal system (Davachi & Wagner, 2002; Jackson & Schacter, 2004; Meltzer & Consta ble, 2005). The behavioral data from the current study also provide information about the pr ocesses involved in relational me mory. Groups achieved fairly comparable levels of learning during training for the overlapping pairs, suggesting that the pairs were encoded to a similar degree across groups. On the test, patients wi th right ATL showed a trend towards poorer performance on the condition that tested me mory for overlapping pairs and showed significantly poorer perf ormance than controls on the condition requiring TI across the overlapping pairs. This pattern of performance suggests that consolidation and retrieval of relational information was difficult for patients w ith right ATL. In particular, the ability to flexibly use retrieved information and apply it to solve a novel problem was impaired. It was hypothesized that lateral ity effects would be observed in the current study, such that TI for nonverbal information would be more diffi cult for patients with right ATL than patients with left ATL. While the task did not clearly di scriminate side of surgery, patients with right ATL did perform more poorly than controls. Pa st research on TI has not been designed to address questions of laterali ty, and hemispheric lateralizat ion observed during functional imaging is not easily interpreted. Heckers and colleagues (2004) demonstrated right anterior hippocampal activation during TI with the paradi gm adapted for use in the current study. However, in a later study using a similar para digm (Zalesak & Hecker s, 2009), lateralization effects were less clear and varied based on the as pect of TI that was studied. Activation was greater in the right hippo campus for pairs that were more adjace nt in the hierarchy, but greater in the left hippocampus for comparisons of pairs th at did not contain end items versus those that

PAGE 64

64 did. The lateralization of TI is deserving of further study using functional imaging and behavioral paradigms. More broadly, TI reflects relational memo ry processing because it examines whether representations that share common elements have been linked in memory to solve new problems (Eichenbaum, 2000). These results provide support for the relational memory account of hippocampal function in a new population, patients with complex partial epilepsy of temporal lobe origin. The findings also suggest a laterali ty effect for TI, such that relational memory processing for visual/nonverbal information is difficult for TLE patients with right hippocampal pathology and side of surgery. Clinical Applications of Transitive Inference A secondary aim of the s tudy was to compare the clinical utility of the TI task to conventional neuropsychological measures. The TI score (i.e., accuracy on inferences for overlapping pairs) was chosen to represent performance because it appeared most sensitive to differential performance in TLE. Material specifi c-predictions link memory for verbal material to the language dominant temporal lobe (usually left) and memory for nonverbal material to the language non-dominant temporal lobe (usually right; Milner, 1975). Accordingly, the TI task was compared to the standard measure of nonverbal memory, the Rey Complex Figure Test. In the current study, results s upport the strong link between impaired verbal memory and left TLE, which has been consistently demonstr ated in the literature (Helmstaedter, 2004; Hermann et al., 1997, Strauss et al., 1995). Scor es from the HVLT-R and the WMS-III Verbal Paired Associates discriminated patients with left ATL from those with right ATL. Both measures assess verbal list learning capacity and showed large effect sizes in discriminating between groups. The largest effect size ( 2 = .30) was observed for the WMS-III Verbal Paired Associates Total score, which s uggests that memory for word pairs as opposed to single words

PAGE 65

65 may be more sensitive to left TLE. This pa ttern would be predicted by relational memory theory. The WMS-III Logical Memory subtest did not demonstrate the capability to discriminate between patients based on side of surgery, alt hough patients with left ATL did perform worse than controls on this task. Conversely, the conventional nonverbal memory measure did not discriminate between groups (right ATL, left ATL, control). This finding is consistent with previous research, which has not found a reliable link between right TLE and nonverbal memory (Barr et al., 1997; Helmstaedter, 2004; Lee et al., 2002 ; Martin et al., 1999). Of note, the epilepsy patient groups and the control group performed in the impair ed to low average range on the Rey Complex Figure Test. The low scores may be indicative of the false lateralizing figural memory performance sometimes observed in left TLE (Loring et al., 2007). The TI task may have more promise as a behavioral indicator of right hippocampal dysfunction than the Rey Complex Figure Test. The motor and constructional abilities involved in figural reproduction may conf ound the assessment of memor y, while the TI task uses a recognition format to bypass this concern. Furthermore, analysis re vealed significant group differences on TI (right ATL < contro l) with a medium effect size ( 2 = .18 to .20). In contrast, the Rey Complex Figure Test did not detect signif icant group differences, and a small effect size ( 2 = .03) was noted for both the immediate and de layed recall conditions of this measure. The clinical utility of these measures was fu rther explored using l ogistic regression and receiver operating characteristic (ROC) curve analysis. Neither measure was a significant predictor of side of surgery, a nd operating characteristics were similar for both measures. Low cut scores are suggested: 75% for the TI score and 30 for the Rey Complex Figure Test Delayed Recall T score. The cut scores were set to optimize positive predictive power and specificity,

PAGE 66

66 which resulted in low sensitivity. For instance, at the 75% cut score on TI, positive predictive power was 89%, but only half of patients with right ATL were detected. Higher cut-offs yielded unacceptably high rates of false positives. Taken together, these findings provide limite d support for our secondary hypothesis that the TI task would be more sensitive to ri ght ATL than the conven tional nonverbal memory measure. The TI task detected differences between right ATL patients and controls, while the Rey Complex Figure Test did not. However, ne ither measure discriminated between epilepsy groups with sufficient accuracy to be used with confidence at the single-patient level. Previous research has not compared TI to re sults of standard neur opsychological testing. Partial correlations controlling for age provided so me evidence of convergent and discriminant validity for the experimental task. All significan t correlations were in the small to moderate range ( r = .3 to .4). TI showed a moderate correlation with an estimate of intellectual functioning, suggesting that it ta ps some general cognitive abilities. TI also showed modest correlations with the Rey Complex Figure Test re call measures, which provides some degree of convergent validity. Both transitive and non-transitive inference scores were associated with a measure of contextual, verbal r ecall and Trail Making Test Part B. The relationships between TI and other measures of verbal memory, language, attention, and executi ve function were not significant, which provides evidence of discriminant validity. N on-transitive inference, however, showed low correlations with several measures of attention and execu tive function. This relationship was unexpected and may s uggest a frontal lobe contribution. Limitations and Future Directions A limitation of the current study is the relativel y sm all sample size that may have resulted in some of the analyses being underpowered. In light of the limited pow er, several exploratory analyses were conducted that were strictly speaking, not statistica lly justified using the models

PAGE 67

67 that were generated. The initial power analysis was based on the one published study of TI in a clinical population. This study compared indi viduals with schizophrenia to nonpsychiatric controls using nearly identical stimuli and a comp arable paradigm (Titone et al., 2004). The TI effect (d = .80) was large (Cohen, 1988). A power analysis (power = .8 0) revealed that 63 participants would be needed to detect a large effect in an omnibus three-group test using an alpha level of .05. Therefore, an optimal sample size would have included at least 20 participants per group. While the control and right ATL groups were close to this target, greater difficulty was encountered in recruiti ng patients who had undergone left ATL. One methodological concern is the verbalizability of stimuli in the task. Nonverbal tests may be contaminated by verbal encoding (Barr et al., 1997). Studies that support this idea have shown an association between nonverbal memory and left hippocampal volume in TLE (Griffith et al., 2003; McConley et al., 2008 ). In another study, patients w ith right TLE showed memory impairment for designs only when their comp lexity exceeded verbal learning capacity (Helmstaedter et al., 1995). In the current study, all participants were questioned about the strategies they used to complete the TI task. So me participants reported applying verbal labels to the stimulus patterns (e.g., pla id), while others reported focu sing on the visual patterns and denied using any verbal strategies. In future st udies, it will be essential to generate stimuli that are difficult to verbally label in order to develop a more pur e measure of nonverbal relational memory. In addition to considering new stimuli, the TI task could potentially benefit from other revisions as well. First, it c ould be beneficial to program the task to ensure that a specific learning criterion is met during training (e.g., 80%). The advantag e to this procedure would be ensuring that all partic ipants achieve a minimal level of learning, which is important when

PAGE 68

68 evaluating neurologically impaired populations. Secondly, on the test, 52% of participants achieved greater than 90% correct on the TI condition. This ceili ng effect suggests that a more difficult task could yield greater information about the range of performance. Designing a more difficult TI task could be easily accomplished by creating a larger hierar chy of overlapping pairs that consists of more than five items. This would include more true TI pairs, like B>D, which do not contain end items. Also, condensing the task to focus only on th e overlapping pairs may optimize clinical utility and brevity. Results showed that condi tions with overlapping pairs tended to be more difficult, as evidenced by lower accuracy scores and slower reaction times. The paradigm in the current study included conditions to demonstr ate dissociations between transitive vs. nontransitive inferential problems and between overla pping vs. non-overlapping pairs. This type of paradigm may be more useful in neuroimaging studies in which it is critical to design tasks that highlight functional dissociations between neuroana tomical regions. In contrast, in a clinical study, the information provided by the additional conditions with non-overlapping pairs is less useful. This study focused on a nonverbal task, given the dearth of behavioral measures sensitive to language non-dominant hippocampal function. However, the study design did not permit a full evaluation of material specificity. A verbal TI analogue would be needed to fully evaluate the material-specific model of memory. The auth or developed a verbal TI task using non-words as stimuli and piloted this task with healthy adults and with several epilepsy patients. Individuals tended to perform more poorly on th e verbal task and described it as more difficult. While the nonverbal task may have permitted dual encoding resulting in better performance, the verbal task may have facilitated strictly ve rbal encoding. Because of difficu lties obtaining adequate pilot

PAGE 69

69 data on the verbal measure, it was eliminated fr om the study. In the future, it will be important to refine this methodology to develop a verbal TI analogue of a similar level of difficulty, in order to facilitate comparison and to evaluate laterality effects. Lastly, the current study evaluated TI in patients who had undergone ATL. Examining the paradigm in post-surgical epilepsy patients is an in itial step. The next step would be to evaluate TI in pre-surgical epilepsy patients to determin e whether the paradigm s hows sensitivity to right temporal dysfunction in this patient group. The c linical implications in this patient group are more relevant, as the eventual applied goal of this research is to develop more effective presurgical tools to assist in the diagnosis of lo calized/lateralized seizure onset and to predict the likelihood of post-surgical decline in cognitive func tion (Helmstaedter, 2004). Conclusion The current study is the first to examine TI in TLE. The study showed that TI for visual inform ation is difficult for patients who underw ent right ATL. Results provide some limited support for the material-specificity of TI. Mo re broadly, the findings suggest a link between hippocampal function and TI, which is consistent with the relational memory c oncept. In light of the lack of behavioral measures sensitive to language non-dominant hippocampal function, the TI paradigm may have clinical promise. In futu re research, it would be beneficial to continue studying TI capacities in the TLE population, using some of the modifications suggested in preceding sections and expanding the work to include pre-surgical epilepsy patients.

PAGE 70

70 LIST OF REFERENCES Abrahams S., Morris, R. G., Polkey, C. E., Jaro sz, J. M., Cox, T. C., Graves, M., et al. (1999). Hippocampal involvement in spatial and workin g memory: A structural MRI analysis of patients with unilateral mesial temporal lobe sclerosis. Brain and Cognition, 41 39. Aggleton, J. P., & Brown, M. W. (1999). Epis odic memory, amnesia, and the hippocampalanterior thalamic axis. Behavioral and Brain Sciences, 22 425. Army Individual Test Battery. (1944). Manual of Directions and Scoring Washington, DC: W ar Department, Adjutant Generals Office. Barr, W. B., Chelune, G. J., Hermann, B. P., Lori ng, D. W., Perrine, K., Strauss, E., et al. (1997). The use of figural reproduction tests as meas ures of nonverbal memory in epilepsy surgery candidates. Journal of the International Neuropsychological Society, 3 435. Barr, W., Morrison, C., Zaroff, C., & Devinsky, O. (2004). Use of the Brief Visuospatial Memory Test-Revised (BVMT-R) in neuropsychological evaluation of epilepsy surgery candidates. Epilepsy & Behavior, 5 175. Baxendale, S. A., Thompson, P. J., & Paesschen, W. V. (1998). A test of spatial memory and its clinical utility in the pre-surgical investigation of temporal lobe epilepsy patients. Neuropsychologia, 36, 591. Benedict, R. H. (1997). Brief Visuospatial Memory Test-Revised Odessa, FL: Psychological Assessment Resources. Brandt, J. & Benedict, R. H. B. (2001). Hopkins Verbal Learning Test-Revised Odessa, FL: Psychological Assessment Resources. Breier, J. I., Plenger, P. M., Castillo, R., Fuchs, K., Wheless, J. W., Thomas, A. B., et al. (1996). Effects of temporal lobe epile psy on spatial and figural aspects of memory for a complex geometric figure. Journal of the International Neuropsychological Society, 2 535. Brown, M. W., & Aggleton, J. P. (2001). Recognition memory: What are the roles of the perirhinal cortex and hippocampus? Nature Reviews Neuroscience, 2 51. Chelune, G. J. (1995). Hippocampal adequacy versus functional reserve: Predicting memory functions following temporal lobectomy. Archives of Clinical Neuropsychology, 10 413 432. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates. Cohen, N. J., Poldrack, R. A., & Eichenbaum, H. (1997). Memory for items and memory for relations in the procedural/d eclarative memory framework. Memory, 5, 131.

PAGE 71

71 Cohen, N. J., Ryan, J., Hunt, C., Romine, L ., Wszalek, T., & Nash, C. (1999). Hippocampal system and declarative (relational) memo ry: Summarizing the data from functional neuroimaging studies. Hippocampus, 9, 83. Davachi, L., & Wagner, A. D. (2002). Hippocampal contributions to episodic encoding: Insights from relational and item-based learning. Journal of Neurophysiology, 88, 982. Davis, R. N., Andresen, E. N., Witgert, M. E., & Breier, J. I. (2006). Is basic memory structure invariant across epilepsy patient subgroups? Journal of Clinical and Experimental Neuropsychology, 28 987. Devito, L. M., Kanter, B. R., & Eichenbaum, H. (in press). The hippocampus contributes to memory expression during tran sitive inference in mice. Hippocampus Dusek, J. A., & Eichenbaum, H. (1997). The hippocampus and memory for orderly stimulus relations. Proceedings of the National Academy of Sciences U S A, 94, 7109. Eichenbaum, H. (2000). A cortical-hippo campal system for declarative memory. Nature Reviews Neuroscience, 1 41. Eichenbaum, H. (2004). Hippocampus: Cognitive processes and neural representations that underlie declarative memory. Neuron, 44, 109. Engel, J., Jr. (1996). Introduction to temporal lobe epilepsy. Epilepsy Research, 26 141. Engel, J., Jr. (2001). Mesial temporal l obe epilepsy: What have we learned? Neuroscientist, 7 340. Frank, J., & Landeira-Fernandez, J. (2008). Co mparison between two scoring systems of the Rey-Osterrieth Complex Figure in left and right temporal lobe epileptic patients. Archives of Clinical Neuropsychology, 23 839. Fortin, N. J., Agster, K. L., & Eichenbaum, H. B. (2002). Critical role of the hippocampus in memory for sequences of events. Nature Neuroscience, 5 458. Giovanello, K. S., Schnyer, D. M ., & Verfaellie, M. (2004). A crit ical role for the anterior hippocampus in relational memory: Evidence from an fMRI study comparing associative and item recognition. Hippocampus, 14 5. Glosser, G., Cole, L., Khatri, U., DellaPietr a, L., & Kaplan, E. (2002). Assessing nonverbal memory with the Biber Figure Learning Test-Ext ended in temporal lobe epilepsy patients. Archives of Clinical Neuropsychology, 17 25. Goodglass, H. & Kaplan, E. (2000). Boston Naming Test-II Philadelphia: Lippincott, Williams, & Wilkins. Gooding, P. A., Mayes, A. R., & van Eijk, R. (20 00). A meta-analysis of indirect memory tests for novel material in organic amnesics. Neuropsychologia, 38 666.

PAGE 72

72 Griffith, H. R., Pyzalski, R. W ., O'Leary, D., Magnotta, V., Bell, B., Dow, C., et al. (2003). A controlled quantitative MRI vol umetric investigation of hi ppocampal contributions to immediate and delayed memory performance. Journal of Clinical and Experimental Neuropsychology, 25 1117. Hannula, D. E., Tranel, D., & Cohen, N. J. (2006). The long and the short of it: Relational memory impairments in amnesia, even at short lags. The Journal of Neuroscience, 26 8352. Heaton, R. K. (1981). Wisconsin Card Sorting Test (WCST) Odessa, FL: Psychological Assessment Resources. Heckers, S., Zalesak, M., Weiss, A. P., Ditma n, T., & Titone, D. (2004). Hippocampal activation during transitive inference in humans. Hippocampus, 14, 153. Helmstaedter, C. (2004). Neuropsychologi cal aspects of epilepsy surgery. Epilepsy & Behavior, 5(Suppl. 1), S45. Helmstaedter, C., Gleissner, U., Di Perna, M., & Elger, C. E. (1997). Relational verbal memory processing in patients with temporal lobe epilepsy. Cortex, 33, 667. Helmstaedter, C., & Kurthen, M. (2001). Memory and epilepsy: Characte ristics, course, and influence of drugs and surgery. Current Opinion in Neurology, 14 211. Helmstaedter, C., Pohl, C., & Elger, C. E. (1995). Relations between verbal and nonverbal memory performance: Evidence of confounding e ffects particularly in patients with right temporal lobe epilepsy. Cortex, 31 345. Hermann, B. P., Seidenberg, M., Dow, C., Jones, J., Rutecki, P., Bhattach arya, A., et al. (2006). Cognitive prognosis in chronic temporal lobe epilepsy. Annals of Neurology, 60, 80. Hermann, B. P., Seidenberg, M., Schoenfeld, J., & Davies, K. (1997). Neuropsychological characteristics of the syndrome of mesial temporal lobe epilepsy. Archives of Neurology, 54, 369. Jackson, O., & Schacter, D. L. (2004). Encoding activity in anterior medial temporal lobe supports subsequent associative recognition. Neuroimage, 21, 456. Jones-Gotman, M. (1991). Localization of lesions by neuropsychological testing. Epilepsia, 32(Suppl. 5), S41. Kneebone, A. C., Lee, G. P., Wade, L. T., & Lo ring, D. W. (2007). Rey Complex Figure: figural and spatial memory before and after tem poral lobectomy for intractable epilepsy. Journal of the International Neur opsychological Society, 13 664. Koehler, S., Danckert, S., Gati, J. S., & Menon, R. S. (2005). Novelty responses to relational and non-relational information in the hippocampus and parahippocampal region: A comparison based on event-related MRI. Hippocampus, 15, 763.

PAGE 73

73 Kroll, N. E., Knight, R. T., Metcalfe, J., Wolf, E. S., & Tulving, E. (1996). Cohesion failure as a source of memory illusions. Journal of Memory and Language, 35 176. Lee, T. M., Yip, J. T., & Jones-Gotman, M. (2002) Memory deficits after resection from left or right anterior temporal lobe in humans: A meta-analytic review. Epilepsia, 43, 83. Libben, M., & Titone, D. (2008). The role of awareness and working memory in human transitive inference. Behavioural Processes, 77, 43. Loring, D. W. (1997). Neuropsychologi cal evaluation in epilepsy surgery. Epilepsia, 38 (Suppl. 4), S18. Loring, D. W., Barr, W., Hamberger, M., & Helmstaedter, C. (2007). Neuropsychology evaluation adults. In T. A. Pe dley & J. Engel, Jr. (Eds.), Epilepsy: A Comprehensive Textbook (2nd ed.) (pp. 1057). Philadelphia: Lippincott, Williams, & Wilkins. Loring, D. W., Lee, G. P., & Meador, K. J. ( 1988). Revising the Rey-Osterrieth: Rating right hemisphere recall. Archives of Clinical Neuropsychology, 3 239. Martin, R. C., Hugg, J. W., Roth, D. L., Bilir, E. Gilliam, F. G., Faught, E., et al. (1999). MRI extrahippocampal volumes and visual memory: Correlations independent of MRI hippocampal volumes in temporal lobe epilepsy patients. Journal of the International Neuropsychological Society, 5 540. Mayes, A. R., Holdstock, J. S., Isaac, C. L., Montaldi, D., Grigor, J., Gummer, A., et al. (2004). Associative recognition in a patient with selective hippocampal lesions. Hippocampus, 14, 763. McConley, R., Martin, R., Palmer, C. A., Kuzniecky, R., Knowlton, R., & Faught, E. (2008). Rey Osterrieth complex figure test spatial and figural scoring: relations to seizure focus and hippocampal pathology in patients with temporal lobe epilepsy. Epilepsy and Behavior, 13 174. McGonigle, B., & Chalmers, M. (1977). Are monkeys logical? Nature, 267, 694. Meltzer, J. A., & Constable, R. T. (2005). Activ ation of human hippocampal formation reflects success in both encoding and cued recall of paired associates. Neuroimage, 24, 384. Meyers, J. E., & Meyers, K. R. (1995). Rey Complex Figure Test and Recognition Trial manual Odessa, FL: Psychological Assessment Resources. Milner, B. (1975). Psychological aspects of focal epilepsy and its neurosurgical management. Advances in Neurology, 8 299. Motamedi, G. K., & Meador, K. J. (2004). Antiepileptic drugs and memory. Epilepsy and Behavior, 5 435.

PAGE 74

74 Nagode, J. C., & Pardo, J. V. (2002). Human hippo campal activation during transitive inference. Neuroreport, 13, 939. OBrien, C. E., Bowden, S. C., Bardenhagen, F. J., & Cook, M. J. (2003). Neuropsychological correlates of hippocampal and rhinal cortex volumes in patients with mesial temporal sclerosis. Hippocampus, 13 892. Oldfield, R. C. (1971). The assessment and an alysis of handedness. The Edinburgh Inventory. Neuropsychologia, 9, 97. Olson, I. R., Page, K., Moore, K. S., Chatterjee, A., & Verfaellie, M. (2006). Working memory for conjunctions relies on the medial temporal lobe. The Journal of Neuroscience, 26 4596. O'Reilly, R. C., & Rudy, J. W. (2001). Conjunctiv e representations in learning and memory: Principles of cortical and hippocampal function. Psychological Review, 108 311. Preston, A. R., & Gabrieli, J. D. (2002). Different functions for different medial temporal lobe structures? Learning and Memory, 9 215. Preston, A. R., Shrager, Y ., Dudukovic, N. M., & Gabrieli, J. D. (2004). Hippocampal contribution to the novel use of relationa l information in declarative memory. Hippocampus, 14, 148. Prince, S. E., Daselaar, S. M., & Cabeza, R. (2005). Neural correlates of relational memory: Successful encoding and retrieval of semantic and perceptual associations. Journal of Neuroscience, 25 1203. Psychological Corporation. (1999). Wechsler Abbreviated Scal e of Intelligence (WASI) San Antonio, TX: Author. Reitan, R. M. (1958). Validity of th e Trail Making Test as an indi cator of organic brain damage. Perceptual and Motor Skills, 8 271. Rickard, T. C., & Grafman, J. (1998). Losing their configural mind: Amnesic patients fail on transverse patterning. Journal of Cognitive Neuroscience, 10 509. Rombouts, S. A., Machielsen, W. C., Witter, M., Barkhof, F., Lindeboom, J., & Scheltens, P. (1997). Visual association encoding activates the medial temporal lobe: A functional magnetic resonance imaging study. Hippocampus, 7, 594. Ryan, J. D., Moses, S. N., & Villate, C. (2009) Impaired relational organization of propositions, but intact transitive inference, in aging: Im plications for understanding neural integrity. Neuropsychologia, 47, 338. Savage, G. R., Saling, M. M., Davis, C. W., & Berkovic, S. F. (2002). Direct and indirect measures of verbal relational memory following anterior temporal lobectomy. Neuropsychologia, 40, 302.

PAGE 75

75 Scoville, W. B., & Milner, B. (1957). Loss of r ecent memory after bilate ral hippocampal lesions. Journal of Neurology, Neurosurgery, & Psychiatry, 20 11. Snitz, B. E., Roman, D. D., & Beniak, T. E. ( 1996). Efficacy of the Continuous Visual Memory Test in lateralizing temporal lobe dysf unction in chronic complex-partial epilepsy. Journal of Clinical and Experime ntal Neuropsychology, 18 747. Sperling, R. A., Bates, J. F., Cocchiarella, A. J., Schacter, D. L., Rosen, B. R., & Albert, M. S. (2001). Encoding novel face-name asso ciations: A functional MRI study. Human Brain Mapping, 14 129. Squire, L. R. (2004). Memory systems of the br ain: A brief history a nd current perspective. Neurobiology of Learning and Memory, 82 171. Staresina, B. P., & Davachi, L. (2006). Diffe rential encoding mechan isms for subsequent associative recognition and free recall. The Journal of Neuroscience, 26 9162. Stark, C. E., Bayley, P. J., & Squire, L. R. (2002). Recognition memory for single items and for associations is similarly impaired foll owing damage to the hippocampal region. Learning and Memory, 9, 238. Stark, C. E., & Squire, L. R. (2003). Hippocampa l damage equally impairs memory for single items and memory for conjunctions. Hippocampus, 13, 281. Strauss, E., Loring, D., Chelune, G., Hunter, M., Hermann, B., Perrine, K., et al. (1995). Predicting cognitive impairment in epilepsy: Findings from the Bozeman epilepsy consortium. Journal of Clinical and Expe rimental Neuropsychology, 17 909. Stroup, E., Langfitt, J., Berg, M., McDermott, M ., Pilcher, W., & Com o, P. (2003). Predicting verbal memory decline following an terior temporal lobectomy (ATL). Neurology, 60 1266. Titone, D., Ditman, T., Holzman, P. S., Eich enbaum, H., & Levy, D. L. (2004). Transitive inference in schizophrenia: Impairment s in relational memory organization. Schizophrenia Research, 68 235. Trahan, D. E., & Larrabee, G. J. (1988). Continuous Visual Memory Test Odessa, FL: Psychological Assessment Resources. Trenerry, M. R. (1996). Neuropsychologic asse ssment in surgical treatment of epilepsy. Mayo Clinic Proceedings, 71 1196. Trenerry, M. R., Westerveld, M., & Meador K. J. (1995). MRI hippocampal volume and neuropsychology in epilepsy surgery. Magnetic Resonance Imaging, 13 1125. Turriziani, P., Fadda, L., Caltagirone, C., & Ca rlesimo, G. A. (2004). Recognition memory for single items and for associat ions in amnesic patients. Neuropsychologia, 42 426.

PAGE 76

76 Van der Jeugd, A., Goddyn, H., Laeremans, A., Ar ckens, L., DHooge, R., & Verguts, T. (2009). Hippocampal involvement in the acquisition of relational associati ons, but not in the expression of a transitive inference task in mice. Behavioral Neuroscience, 123 109. von Fersen, L., Wynne, C. D., Delius, J. D., & Staddon, J. E. (1991). Transitive inference formation in pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 17, 334. Wechsler, D. (1997a). Wechsler Adult Intelligence ScaleIII manual San Antonio: The Psychological Corporation. Wechsler, D. (1997b). Wechsler Memory Scale-III manual San Antonio: The Psychological Corporation. Weniger, G., Boucsein, K., & Irle, E. (2004). Im paired associative memory in temporal lobe epilepsy subjects after lesions of hippocampus, parahippo campal gyrus, and amygdala. Hippocampus, 14, 785. Yang, J., Weng, X., Guan, L., Kuang, P., Zhang, M., Sun, W., et al. (2003). Involvement of the medial temporal lobe in pr iming for new associations. Neuropsychologia, 41 818. Yonelinas, A. P., Hopfinger, J. B., Buonocore, M. H., Kroll, N. E., & Baynes, K. (2001). Hippocampal, parahippocampal and occipital-temporal contributions to associative and item recognition memory: an fMRI study. Brain Imaging, 12 359. Zalesak, M., & Heckers, S. (2009). The role of the hippocampus in tr ansitive inference. Psychiatry Research: Neuroimaging, 172 24.

PAGE 77

77 BIOGRAPHICAL SKETCH Ms. Barker grew up in Bristol, Te nnessee. Sh e graduated from the University of the South at Sewanee in 2000 with a Bachelor of Scien ce degree in Psychology and a minor in German. After graduating from college, she studied psyc hology in Munich, Germany with a Fulbright Fellowship. She then worked for two years as a Re search Assistant at the Yale University Child Study Center before beginning graduate school in Clinical Psychology at the University of Florida. She earned her Master of Science degree from the Univ ersity of Florida in 2005. Her graduate work has focused on neuropsychology in epilepsy, traumatic brain injury, and dementia. She is currently attending the Char leston Consortium Internship Program in the Neuropsychology Track.