An Empirical Test of Two Opposing Theoretical ... - Science Direct

15 downloads 0 Views 2MB Size Report
of prefrontal patients on the Self-Ordered Pointing Task (SOPT), (Kolb. & Wishaw, 1990; Milner ...... The life of the mind. Hillsdale: Lawrence Erlbaum Associates.
BRAIN

AND

COGNITION

19,

48-71 (1992)

An Empirical Test of Two Opposing Theoretical Models of Prefrontal Function SYLVIE DAIGNEAULT,

CLAUDE

M. J. BRAZEN, AND HARRY A. WHITAKER

Laboratoire a2 Neurosciences Cognitives and Dkpartement de psychologie, Universitk du Qukbec d Montreal, C. P. 8888, Sum. A, Montrkal, Quebec, Canada, H3C 3P8 The purpose of this study was to compare the validity of two models which contrast with each other in the manner in which they integrate neuropsychological tests into distinct prefrontal constructs. The first prefrontal model consists of five distinct functional constructs drawn from human clinical neuropsychology. The second model, elaborated by Goldman-Rakic, is based primarily on monkey research and postulates a basic prefrontal function, “on-line representational memory,” which guides behavior in the absence of, or despite discriminative environmental stimuli. In the latter model, distinct prefrontal functional constructs are primarily defined in terms of various types of representational memory involved in specific tasks. Eleven “prefrontal” measures were obtained from 259 normal adults, stratified for age, education, and sex. Confirmatory factor analyses revealed that the Goldman-Rakic model “fit” the data better than the model derived from human clinical neuropsychology, while several constructs commonly used in human neuropsychology were refuted. It was concluded that new research on braindamaged humans with a view to understanding prefrontal function might benefit from using the Goldman-Rakic model as a starting point. o 1992 Academic press, IX.

Much research in human neuropsychology seeks to establish correspondences between cerebral lesion loci and performance on tests. These correspondences are further interpreted as localized brain functions. This is how group studies of patients with prefrontal lesions have led to the formulation of “prefrontal” functions. However, the study of human prefrontal functions is subject to several methodological difficulties. Patients with discrete prefrontal lesions are few, such that studies of human prefrontal cases have rarely been able to separate dorsolateral from orbital or medial effects. In addition, prefrontal measures seem to often reflect contributions from other cerebral areas. These problems compel researchers to analyze the different processes involved in each task and to infer functions which might be impaired specifically because of prefrontal damage. Functions dependent upon distinct areas within prefrontal cortex must be even more remotely inferred. Empirical evidence of convergent 48 0278-2626192$5.00 Copyright 8 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.

OPPOSING PREFRONTAL

FUNCTION MODELS

49

validity of measures supposed to reflect a unitary function, and of divergent validity of measures supposed to reflect distinct functions, is practically nonexistent. Concurrent with the research on humans with prefrontal lesions, an ensemble of neuroanatomical, neurophysiological, neuropharmacological, and neurobehavioral findings pertaining to prefrontal cortex in the monkey constitutes an alternative approach. Goldman-Rakic, an architect of this latter approach, has recently proposed an integrated model of prefrontal function(s) which she believes can be generalized to humans (GoldmanRakic, 1987). The purposes of the present study were (1) to test the validity of five prefrontal functional constructs judged by us to be among the most clearly articulated, and frequently cited, in the human neuropsychology literature, and (2) to test the relative validity of Goldman-Rakic’s model. A Prefrontal Model Drawn from Human Neuropsychology

The five prefrontal functional constructs drawn from human neuropsychology, and associated measures, which were subjected to tests of validity, were the following. 1. Planning (or the elaboration of strategy) and execution of sequences of planned responses. This theoretical functional construct has been formulated by several authors as an interpretation of impaired performance of prefrontal patients on the Self-Ordered Pointing Task (SOPT), (Kolb & Wishaw, 1990; Milner, Petrides & Smith, 1985; Petrides and Milner, 1982; Stuss & Benson, 1986; Wiegersma, Van der Scheer & Human, 1990) and of the high number of wrong entries on the Porteus Maze Test by similar patients (Lezak, 1983; Porteus & De Monbrun Kepner, 1944; Stuss & Benson, 1984, 1986; Walsh, 1978). 2. Self-regulation of behavior in response to environmental contingencies (including one’s own errors). This construct has been applied to high rates

of perseverative errors on the Wisconsin Card Sorting Test (WCST) in prefrontal patients who do not modify their course of action despite repeated error feedback (Jones-Gotman, Ptito, & Zatorre, 1984; Luria, 1973; Milner & Petrides, 1984) and to high rates of repeated errors (perseverative errors) on the Porteus Maze test (Petrie, 1949; Porteus & Peters, 1947; Walsh, 1978). 3. Maintenance of a nonautomatic cognitive or behavioral set. This construct has been formulated to explain three types of error scores typical of prefrontal patients: interference errors on the Stroop Test (Lezak, 1983; Stuss & Benson, 1984), category breaks on the WCST (Stuss et al., 1983; Stuss & Benson, 1986), and alternation errors on the Trail Making Test, form B (Reitan, 1958; Stuss & Benson, 1986). 4. Spontaneity/sustained mental productivity. This construct has been linked to prefrontal patients’ difficulties on verbal fluency and design

50

DAIGNEAULT,

BRAtiN,

AND

WHITAKER

fluency tasks (Brown, 1988; HCcaen & Ruel, 1981; Jones-Gotman, Ptito, 8c Zatorre, 1984; Holst & Vilkki, 1988; Miceli, Caltagirone, Gainotti, Masullo, & Siveri, 1981; Ramier & Hecaen, 1970; Walsh, 1978). 5. Spatiotemporal segmentation and organization of events. This theoretical construct, termed “contextual chunking” by Schacter (1987) has been invoked to explain prefrontal patients’ impairment on tasks of recency judgements against a background of spared recognition memory (Kolb & Wishaw, 1990; Ladavas, Umilta & Provinciali, 1979; Milner, Petrides & Smith, 1985; Seron, 1978; Stuss & Benson, 1984) and confusion errors on different lists of recalled items (Luria, 1976, 1980; Schacter, 1987; Stuss & Benson, 1986). Goldman-Rakic’s

Model of Prefrontal

Function

Goldman-Rakic (1987) postulated that prefrontal cortex receives sensory and mnemonic representations of reality as well as symbolic representations (e.g., concepts, plans) which have been elaborated in other cerebral areas. These are kept activated (“on line”) by prefrontal cortex in “representational memory” long enough for this live memory to modulate behavior appropriately despite the absence of external contingencies or despite the presence of external task-irrelevant “discriminative” stimuli. Different prefrontal areas are postulated to house different representational memory units which are related to each other anatomically and functionally. Some of these units are proposed to control highly specific motor responses (eye movements, verbal output), while others are postulated to control various other motor centers. Any one prefrontal area is assumed to exert inhibitory as well as excitatory influence on the relevant motor systems. Discrete prefrontal lesions are understood to hinder specific representational memory units resulting in specific difficulties in the regulation of behavior. The relative validity of this model will be statistically estimated in this study. This estimate must be considered relative because there exist no tests developed in accordance with the model. However Goldman-Rakic (1987) explicitly integrated most of the above-cited test measures into her own model in an innovative manner. To the best of our knowledge, these measures would correspond, in Goldman-Rakic’s model, to four distinct functional constructs: 1. Regulation of manual responses by verbal prefrontal representational memory. On the WCST, the subject must guide his or her motor responses (card assortment) by means of an “on-line” verbal recall of the three classification principles despite absence of discriminative stimuli (which are not in the cards) (Goldman-Rakic, 1987, p. 379). According to this model, category breaks and perseverative errors would result from the same functional alteration (p. 379). On form B of the Trail Making Test, the motor response (pencil tracing) must also be constantly modulated

OPPOSING PREFRONTAL

FUNCTION MODELS

51

by “on-line” verbal memory of the instructions (“a letter, a number, a letter. . .” ), which if impaired would result in alternation errors. 2. Regulation of verbal responses by verbal prefrontal representational memory. Verbal fluency tasks engage “on-line” verbal memory of the

instruction, associated accessto the verbal lexicon and regulation of verbal output. The interference condition of the Stroop test primarily requires constant regulation of verbal responses by a verbal instruction, despite presence of distracting discriminative stimuli-in this case the semantic meanings of the stimulus words (p. 380). On a recall task of several word lists, it is necessary to judge the relative recency and/or the association of past stimuli within prefrontal representational memory. Since the task used in the present study required oral responses, the prefrontal representational memory must also modulate verbal output.rX2 3. Regulation of manual responses by prefrontal visuospatial representational memory. On the Porteus Maze Test the motor response is mod-

ulated by “on-line” representation of dead-end passages (visuospatial memory), preventing entry into them, and by the configuration of a previous trial (visuospatial memory), preventing repetition of a wrong entry (perseverative error). On design fluency tasks, on-line representation of diverse visuospatial configurations is necessary to modulate the response, not only to execute novel ones but also to prevent perseveration (p. 380). 4. Regulation of simple responses by prefrontal representational memory on tasks which are both verbal and visual. On the SOPT, regulation of

behavior depends in part on representational memory of preceding responsesso as not to point twice to a same stimulus (p. 388). This prefrontal process is both visual and verbal since stimulus organization strategies and some aspectsof the stimulus forms themselves (waves, symbols, roofs, etc.) are verbalizeable. The SOPT also depends on verbal strategies of response organization including anticipated responses (pp. 380-382). Goldman-Rakic postulates that judgements of relative recency require regulation of responses by representational memory of past events. The recency memory task used in the present study also would involve verbal and visual representational memory because the stimuli used were highly nameable. So, for both these tasks it is supposed that two anatomically ’ Prefrontal regulation of a verbal response can dissociate from prefrontal regulation of another type of motor response, even if the representational memory leading to the response is the same, because each system is constrained by its output connectivity (p. 379). ’ Goldman-Rakic reports that a principal sulcus lesion in the monkey can impair performance as much on a task with a long response delay as with a short response delay (p. 388). These two conditions are different in that the first involves the hippocampal system, whereas the second does not. This suggeststhat the same prefrontal area may be specialized for a modality of representational memory (ex: visuospatial) regardless of the memory load involved in the task.

52

DAIGNEAULT,

BRAtiN,

AND WHITAKER

distinct systems of representational memory (verbal and visual) are required to collaborate to satisfactorily execute the task. The first (neuropsychological) model is based on data obtained from brain damaged humans. It is eclectic, unintegrated, and its five factors represent discrete psychological inferences. These factors were formulated, at the outset, in terms of distinct psychological constructs which were then attached to vaguely defined prefrontal brain areas. The Goldman-Rakic model, on the other hand, is integrated; it was formulated, at the outset, in terms of distinct brain-behavior relationships described in the animal literature. These were then interpreted in terms of one unique and parsimonious construct (representational memory) which was then fractionated into input and output modalities. To test the validity of these two prefrontal models, confirmatory factor analyses were planned. Such analyses were judged more appropriate for this purpose than exploratory factor analyses because they afford a direct test of explicit models such as the ones outlined above. Contrary to exploratory factor analysis, confirmatory factor analysis is less subject to self-serving manipulation of analytical parameters (axis rotation method, factorial correlation constraints, factor retention criteria) and post hoc interpretive arbitrariness. The prefrontal measures selected for this study have all been validated in terms of the criterion of being sensitive to prefrontal lesions. However, we suggest that analyses of cons~rrucfvalidity can be usefully performed using data from normal subjects. Short of large samples of brain-damaged patients, marked stratification of sampling of many normal subjects will assure a sufficient variance on most measures. Variability is, of course, a prerequisite of any attempt to infer links between specific normal behaviors and specific brain functions. It is now well established that, though large samples of brain-damaged subjects differ in performance on individual neuropsychological measures from normal samples, the basic relationships between the measures (i.e., the factor structure) is the same in both types of sample (California Verbal Learning Test: Delis, Freeland, Kramer, & Kaplan, 1988; Wechsler Adult Intelligence Scale: Russell, 1972). In cognitive neuropsychology it is assumed that normal cognitive structure (distinct subsystems or modules) is reflected in, and emerges from, normal brain organization. It follows that symptoms of what could be thought of as theoretically equivalent to brain damage can be observed in normal subjects, each of which possesses modules which function better than others (inter- and intrasubject neurecognitive variability) (Ellis & Young, 1988). METHODOLOGY

Subjects Two hundred and fifty-nine normal adults were retained for the analyses presented here. The sample was stratified for age, education, and sex. Mean age was 38.2 years (SD =

OPPOSING PREFRONTAL

FUNCTION MODELS

53

TABLE 1 DIWRIFTION OF THE SAMPLE Age

Less than 12 years of education 12 to 14 years of education 15 to 17 years of education 18 and more years of education

16-30

31-45

46-65

18 men 18 women 20 men 18 women 15 men 10 women 6 men 4 women

19 men 11 women 8 men 8 women 16 men 12 women 10 men 7 women

11 men 18 women 8 men 6 women 5 men 5 women 3 men 3 women

Note. Thirty-six of these subjects were students at the time of testing. The others (n = 223) had finished their schooling. 13.0, range = 16-65). Mean education was 13.3 years (SD = 3.8, range = 4-23). See Table 1. At the end of the evaluation, a questionnaire was administered which included questions about age, education, spoken languages, consumption of alcohol, drugs and medication, medical history (hospitalizations, surgeries, diseases, psychological, psychiatric or neurological consultation, pharmacotherapies). This resulted in the exclusion of 26 subjects with possible brain dysfunction from the initial cohort of 285 subjects.’

Tests 1. Self-Ordered Pointing Task @OPT) developed by Petrides and Milner (1982). Twelve series of abstract designs were presented in exact accordance with the author’s procedure. 2. Wisconsin Card Sorting Test (WCST). The 128 card version was administered using Heaton’s (1981) procedure. 3. Porteus Maze Test (Porteus, 1965). The four most advanced mazes (XI, XII, XIV, Adult I) were administered using Porteus’s procedure. 4. Verbal Fluency. The Controlled Oral Word Association Test of Benton and Hamsher (1983) (using the letters PFL, as recommended for the French language by Ramier and Hecaen, 1970) was otherwise administered according to the author’s procedure. 5. Design Fluency. The first condition of the procedure published by Jones-Gotman and Milner (1977) was used. 6. Stroop Test. The three conditions were administered in the following order: (a) color denomination of patches, (b) monochromatic reading, and (c) the interference condition, each for 45 set according to Golden’s (1978) procedure. 7. Recency Test. The procedure, similar to that of Corsi (1972), comprised 61 stimulus pages, each containing two representative line drawings (taken from the Peabody Picture Vocabulary Test) to be memorized. There were also 28 probe pages, containing two drawings and a large question mark. In the case of eight of these, only one of the stimuli had been previously presented, while the rest contained two drawings which had both been previously seen. When a probe page occurred, the subject was requested to state whether he/she 3 Exclusion criteria were the following: more than 24 beers, more than 5 bottles of wine, or more than 15 oz of spirits per week; occasional consumption of cocaine, LSD, or psychostimulants; any consultation in neurology, or psychiatry (involving a drug treatment); cranial trauma with hospitalization or any major operation with general anesthesia (ex: cardiac).

54

DAIGNEAULT,

BRAUN, AND WHITAKER

recalled having seen one or both probes (recognition), and in the latter case, which he/she had seen most recently (relative recency judgement) (Daigneault, Bratin, Proulx, & Gilbert, 1988). 8. Verbal lenming with interference procedure. Rey’s (1970) lists A and B (15 words per list) were used in a procedure similar to Lezak’s (1983), comprising four steps: (a) after having heard list A, the subject tried to repeat the words aloud; (b) after a second presentation of list A, the subject again tried to recall the words aloud; (c) after hearing list B, the subject tried to repeat the words of this new list aloud (interference strategy); and (d) the subject was then required to recall the words of list A, without hearing the words anew. 9. Trail Making Test. Forms A and B were administered using Reitan’s (1958) procedure.

The Prefrontal Measures 1. SOPT-E. The number of errors on the Self-Ordered Pointing Task, corresponding to the total number of times any stimulus is pointed to twice within a “series” (See Petrides & Milner, 1982 for details). 2. MAZES-NPE. The number of nonperseverative errors, i.e., the number of times a subject entered a new dead-end path on the Porteus Maze Test. 3. WCST-PE. The number of perseverative errors on the Wisconsin Card Sorting Test as defined by Heaton (1981). 4. MAZES-PE. The number of times a subject entered twice into the same dead-end path on the Porteus Maze Test. 5. STROOP-E. The number of errors (reading the word instead of naming the ink color) on the interference condition of the Stroop Test. 6. WCST-CB. The number of category breaks on the Wisconsin Card Sorting Test. This type of error consisted of one incorrect response after a minimum of three contiguous and nonambiguous correct responses, or after five contiguous correct responses, one of which was nonambiguous. 7. TRAILS-E. The number of alternation errors on form B of the Trail Making Test. 8. V-FLUENCY. The number of correct words on the Controlled Oral Word Association Test. 9. D-FLUENCY. The number of different and unnamable drawings on the first condition of the Design Fluency Test. 10. RECENCY. The proportion of recency judgment errors out of the items explicitly recognized on the Recency Test. 11. CONFUSION. The number of words from list A reported on recall of list B, divided by the number of correctly recalled words, plus, the number of words from list B reported on recall of list A, divided by the number of correctly recalled words on the last step of the Verbal Learning Test.

Confirmatory

Factor Analyses (CFA)

Advantages of CFA include the following: (a) it allows a relatively direct test of the construct validity of a theory or model by measuring statistically the extent to which a postulated factor structure “fits” the actual data, (b) it can quantitatively support comparison of several competing models, and (c) it can orient the analyst toward means by which to improve the original theoretical model. CFA requires that a model be specified in a precise manner. For example, the factor structure must be prespecified in terms of number of factors, factor loadings, and interfactor relatedness. Of course, there is always room for interpretation of how to specify a theory in terms of factor structure (Newby, Hallenbeck, & Embretson, 1983). The softward used in the present study, LISREL-7, offers options for the analysis of ordinal variables, for the production of matrices of such correlations (polychoric) as well

OPPOSING PREFRONTAL

FUNCTION MODELS

55

as for the methods of estimation of “fit” which do not assume (as does the maximum likelihood method) multivariate normality. One such method is the unweighted least squares (ULS) method. Statistical indexes of “fit” are: (a) Chi square and its probability level, which can be validly applied only to multivariate-normal distributions; (b) the adjusted goodness of fit index (AGFI) (it is adjusted for degrees of freedom); (c) the root mean square residual (RMSR); and (d) the coefficient of determination (CD), a general index of the strength of combined correlations of variables with their designated factors. The AGFI and RMSR parameters are not known to be related to any particular distribution, do not require multinormality, and do not yield probability estimates. An AGFI greater than .8 is a generally accepted criterion of good fit, while a RMSR smaller than .l is a generally accepted criterion of sufficient fit (Cole, 1987). Furthermore, the CFA program used here provides several other meaningful indications of poor fit of the data to the model. These are termed “failures of admissibility” and usually result in abortion of the run. They include “nonconvergence to a solution,” “non-positively defined matrices,” and “interfactorial correlations greater than one.” Two statistical indexes provide information about individual variables within the multivariate environment. The first, the multiple-squared correlations, constitutes a measure of the strength of the relation between each variable and its designated factor. The second, the standardized residual covariances, indicates, when their values are greater than two, a specificity deficit of the model, in other words, the model is thought to fail to adequately explain residual covariance (Dillon & Goldstein, 19&t; Joreskog & Sorbom, 1984).

RESULTS

The majority of measures were not normally distributed, and some were practically dichotomous. The purpose of this investigation was to analyze constructs reflected by the various measures rather than to analyze psychometric properties of the measures themselves. Consequently, it was judged appropriate to reduce the psychometric properties of each distribution to a common, ordinal level, denominator. All variables were transformed into ordinally distributed values, limited to a maximum of five. The procedure SAS PROC RANK was used for this purpose. This procedure standardizes the intervals and approximately equalizes the samples within each interval. The measures SOPT-E, WCST-PE, STROOP-E, VFLUENCY, D-FLUENCY, and RECENCY took five values, MAZESNPE took four, WCST-CB and CONFUSION took three, and MAZESPE and TRAILS-E took two. These distributions were formulated so that increasing rank signified better performance (less errors), thus simplifying interpretation of subsequent analyses. All estimates of “fit” for the CFAs reported below were derived by means of the method of unweighted least squares. The matrix of polychoric correlations, calculated by LISREL-7, served as input for all CFAs on the 11 prefrontal measures designed to test the validity of the two models. See Table 2. CFAs of the Human Neuropsychology

Model

The CFA designed to test this model was made as parsimonious as possible. The five factors were prespecified to be determined exclusively

.195***

.074

.292***

.076 .105 .149* .049

.147*

.302*** .154***

2

.234***

.163**

.320*** ,082 .265*** .270*** .200*** .102

3

CORRELATIONS

.291***

.215*** .033

.418*** .618*** 390 .142* .132*

.532***

1

OF POLYCHORK

TABLE 2

.261*** .104

-.072 .420*** ,062

- ,061

.126

4

- .017 .139*

.098 -.140* .210*** .021

5

SERVING AS INPUT TO ALL

6

ON THE

.140*

.125*

.061 .190**

.233***

CFAs

.I08 .215*** .149* -.022

7 8

MEASURES

.280*** .147** .248***

11 PREFRONTAL

.llO

,013

9

.291***

10

Note. After Bonferroni correction (alpha/number of comparisons) for this number of comparisons an individual alpha level of .OOl corresponds to a corrected alpha level of .05 * p < .05. ** p < .Ol. *** p < .ool.

1. SOPI--E 2. MAZES-NPE 3. WCST-PE 4. MAZES-PE 5. STROOP-E 6. WCST-CB 7. TRAILS-E 8. V-FLUENCY 9. D-FLUENCY 10. RECENCY 11. CONFUSION

MATRIX

g

g rl k

5

F c: 3

B s .!!

z 6

OPPOSING PREFRONTAL

57

FUNCTION MODELS

TABLE 3 FACTOR STRUCTURE DESIGNED TO TEST THE HUMAN

SOPT-E MAZES-NPE WCST-PE MAZES-PE STROOP-E WCST-CB TRAILS-E V-FLUENCY D-FLUENCY RECENCY CONFUSION

NEUROPSYCHOLOGY

MODEL

Fl

F2

F3

F4

F5

X X 0 0 0 0 0 0 0 0 0

0 0 X X 0 0 0 0 0 0 0

0 0 0 0 X X X 0 0 0 0

0 0 0 0 0 0 0 X X 0 0

0 0 0 0 0 0 0 0 0 X X

Note. Fl, Planning (or the elaboration of strategy) and execution of sequences of planned responses; F2, self-contingencies (including one’s own errors); F3, maintenance of a nonautomatic cognitive or behavioral set; F4, spontaneity/sustained mental productivity; and F5, spatiotemporal segmentation and organization of events. X, unconstrained factorial weight; 0, factorial weight hxed at 0.

by the variables which have been associated with them in the literature as presented in our introduction above. The weight of each variable onto its designated factor was left unspecified. The interfactorial correlations were left unspecified since a sizeable G factor would be expected for any such cognitive variables in neurologically intact subjects (Wechsler, 1981). This factor structure is schematized in Table 3. This CFA indicated poor fit of the model to the data and it even failed the admissibility test. The correlation between Fl (planning, or the elaboration of strategy, and execution of sequencesof planned responses) and F2 (self-regulation of behavior in response to environmental contingencies, including one’s own errors) was greater than 1. To get some indication of the validity of the remaining factor structure, a new CFA was planned, collapsing Fl and F2. This new CFA therefore tested a 4 factor model wherein the original first two factors were not kept separate. The measures SOPT-E, MAZES-NPE, WCST-PE, and MAZES-PE were all constrained onto a single “nontheoretical” factor which is not purported to correspond to any meaningful prefrontal construct. This CFA passed the admissibility tests and otherwise showed good fit. The AFGl was .952, the RMSR was .055, and the CD was .925. Only one standardized residual covariance (WCST-PE/WCST-CB) attained a value greater than 2. All the above suggestsgood fit. However, one of the remaining factors appeared nevertheless quite weak; the 3rd factor, “maintenance of a nonautomatic cognitive or behavioral set” related weakly to its designated variables. The STROOP-E, WCST-CB, and TRAILS-E measures had squared multiple

58

DAIGNEAULT, BRAtiN, AND WHITAKER TABLE 4 FACTOR STRUCTUREDESIGNED TO TEST THE GOLDMAN&SIC

SOFT-E MAZES-NPE WCST-PE MAZES-PE STROOP-E WCST-CB TRAILS-E V-FLUENCY D-FLUENCY RECENCY CONFUSION

MODEL

Fl

F2

F3

F4

0 0 X 0 0 X X 0 0 0 0

0 0 0 0 X 0 0 X 0 0 X

0 X 0 X 0 0 0 0 X 0 0

X 0 0 0 0 0 0 0 0 X 0

Note. Fl, Regulation of manual responses by verbal prefrontal representational memory; F2, regulation of verbal responses by verbal prefrontal representational memory; F3, regulation of manual responses by prefrontal visuospatial representational memory; F4, regulation of simple responses by prefrontal representational memory on tasks which are both verbal and visual. X, unconstrained factorial weight; 0, factorial weight fixed at 0.

correlations of only .054, .078 and .033, respectively. After Bonferroni correction, STROOP-E was nonsignificantly correlated with WCST-CB and with TRAILS-E and only WCST-CB and TRAILS-E were significantly correlated, at the univariate level. CFAs of the Goldman-Rakic

Model

As in the previous analyses, the Goldman-Rakic model discussed in the introduction was tested parsimoniously with variables assigned exclusively to one predesignated factor, the weights of each variable onto their designated factor left unspecified, and the interfactorial correlation left unspecified. See Table 4. The CFA of the Goldman-Rakic model indicated good fit in all respects. Its AFGI was .915, its RMSR was .080, its CD was .997. However, 4 squared multiple correlations were below .l (STROOP-E, WCST-CB, TRAILS-E, and D-FLUENCY) and 9 of 55 standardized residual covariances were greater than 2. Given that none of these dependent measures have been designed with the intention of testing Goldman-Rakic’s model, it should not be surprising that certain measures correlate weakly with their designated factor or that some of the residual covariance is high. Detailed examination of simple correlations between those measures purported to form separate factors helps in the understanding of the convergent validity of the postulated constructs. In this respect (a) the three measures of the first factor, WCST-PE, WCST-CB, and TRAILSE, were all significantly intercorrelated, at the univariate level, even after

OPPOSING PREFRONTAL

FUNCTION MODELS

59

Bonferroni correction (b) the three measures of the second factor, VFLUENCY, STROOP-E, and CONFUSION, were also all significantly intercorrelated. However, after Bonferroni correction STROOP-E and CONFUSION were no longer significantly correlated, (c) two of three measures of the third factor MAZES-NPE and MAZES-PE were significantly and highly correlated, but the third measure, D-FLUENCY was not correlated with them, and (d) the measures of the fourth factor, SOPTE and RECENCY, were significantly intercorrelated, even after Bonferroni correction. Bentler & Bonett (1980) have suggested that the relative adequacy of a model can be tested by comparing it to a null model that postulates an absence of correlation between the variables. However, because cognitive abilities are often correlated in normal subjects, Moehle, Rasmussen, & Fitzburgh-Bell (1987) have argued that a better way to test the relative adequateness of a model is to compare it to a model where all constraints are the same except that variables are randomly assigned to the factors. We decided to apply both types of tests. The first null model corresponded to the postulate of no correlation between the variables. This CFA revealed very poor fit. The AFGl was only .452 and the RMSR was .219. A second null model corresponded to the postulate of a single factor (G factor) explaining all the variance and covariance. This CFA obtained an AGFl of .892 and a RMSR of .097, indicating acceptable fit. However, there were 15 standardized residual covariances greater than two, indicating that the model was far from accounting for the multivariate space adequately. Random attribution of variables to a CFA model structured the same way as the Goldman-Rakic model was prepared. The three variables assigned to Fl were V-FLUENCY, CONFUSION, and STROOP-E; to F2, MAZES-NPE, MAZES-PE, D-FLUENCY; to F3, WCST-PE, RECENCY, TRAILS-E; and to F4, WCST-CB and SOPT-E. A second similar CFA design assigned the same variables but in inverse order (the first two to F4, the next three to F3, etc). These two “random” CFAs resulted in very poor fit. The first failed the admissibility test because F3 and F4 correlated beyond 1, and the second because two interfactor correlations (Fl/F2 and F2/F4) were greater than 1. Thus, the Goldman-Rakic model, as tested by the CFA structure thought to best represent it, “explained” the 11 prefrontal measures better than two null models and two random models, as well as better than the model derived from the human neuropsychology literature. CFAs Designed to Test the Divergent Validity of the Factor “Spatiotemporal Organization and Segmentation of Events,” Derived from Human Neuropsychology, within a Wider Mnemonic Space

The following CFAs concerned measures corresponding to the fifth factor of the human neuropsychology model, namely, “spatiotemporal

60

DAIGNEAULT,

BRAUN, TABLE

MATRIX

AND

WHITAKER

5

OF POLYCHORIC GXRELATIONS SERVING AS INPUT FOR CFAs ON THE FOUR MEMORY MEASURES

RECENCY CONFUSION RECOGN V-LEARNING Note. After Bonferroni comparisons an individual *** p < .ool.

CONFUSION

.291*** .470*** .310***

.293*** .292***

RECOGN

.359***

correction (alpha/number of comparisons) for this number of alpha level of .008 corresponds to a corrected alpha level of .05.

organization and segmentation of events.” It also addressed two “memory” measures which the human neuropsychology literature would suppose to be relatively independent of prefrontal areas. The nonprefrontal measures were RECOGN, number of nonrecognized stimuli on the recency test and V-LEARNING, number of words recalled from list A on both first trials on the verbal learning test. As had been done for both prefrontal measures, these were transformed into ordinal distributions (again, using SAS PROC RANK). They both took five values and were coded so that increasing rank corresponded to better performance. The matrix of polychoric correlations, serving as input for the LISREL-7 calculation of these CFAs, is presented in Table 5. A first CFA was designed to test the degree of fit of the two “prefrontal” measures (RECENCY and CONFUSION) to a “prefrontal” memory factor, and the two “nonprefrontal” measures (RECOGN and V-LEARNING) to a nonprefrontal memory factor. The weight of each measure on its factor was left unconstrained and was fixed at zero for the other factor. Interfactor correlation was left unconstrained given the now well known “general memory” factor in normal subjects (Delis, Freeland, Kramer, TABLE FACTOR STRUCTURE FOR TESTING THE DIVERGENT “SPATIOTEMPORAL SEGMENTATION AND ORGANIZATION MEMORY FUNCXIONS

RECENCY CONFUSION RECOGN V-LEARNING

6 VALIDITY OF THE PREFRONTAL FUNCTION OF EVENTS” VERSUS OTHER NONFRONTAL

Fl

F2

X X 0 0

0 0 X X

Note. Fl, Prefrontal mnemonic function “Spatiotemporal segmentation and organization of events”; F2, nonprefrontal mnemonic function. X, unconstrained factorial weight; 0, factorial weight fixed at 0.

OPPOSING PREFRONTAL

FUNCTION MODELS

61

& Kaplan, 1988; Wechsler, 1987). The factor structure is schematized in Table 6. This CFA indicated very poor fit; the correlation between Fl (prefrontal mnemonic function) and F2 (nonprefrontal mnemonic function) was greater than one. To better understand these negative results, other alternative constructs were tested with additional CFAs. The first model corresponded to the postulate of no correlation between the four measures. A second model corresponded to the postulate of a single general “memory” factor. A third model corresponded to the postulate of correlation of measures drawn from the same test or comprising the same stimuli (Fl = RECENCY and RECOGN, F2 = CONFUSION and V-LEARNING). The first CFA (noncorrelation) provided poor fit (AGFI = .291, RMSR = .264). However the second CFA (a single memory factor) provided very good fit (AGFI = .991, RMSR = .023, CD = .987). None of the standardized covariance residuals were greater than two, and the smallest squared multiple correlation was .22 (CONFUSION). The third CFA (same test or same stimuli) also indicated very good fit (AGFI = .997, RMSR = .008, CD = .757). None of the standardized covariance residuals were greater than two, and the smallest squared multiple correlation was .26 (CONFUSION). These CFAs further challenge the prefrontal construct “spatiotemporal organization and segmentation of events,” by revealing that two alternative constructs provide much better “fit” within this particular mnemonic space. DISCUSSION Methodological

Considerations

Confirmatory factor analysis for testing models with data from normal subjects stratified for sex, age, and education has numerous advantages. Confirmatory factor analysis provides statistical tests of relative validity of multivariate models and even tolerates nonnormally distributed input (typical in the domain of clinical neuropsychology). This approach is particularly relevant here given that prefrontal functions are recognized by everyone as latent constructs that are not directly measurable. The use of normal subjects for construct validity studies makes it possible to collect large data bases in terms of both the number of variables per subject and the number of subjects. Correlations between the three stratification variables, sex, age, and education, and the eleven prefrontal measures indicated that only age and education were significantly correlated with some of the measures. Taken together, the stratification variables did not explain more than 32% of the variance of the measures as determined by canonical correlation anal-

62

DAIGNEAULT,

BRAijN,

AND

WHITAKER

ysis. This suggests that intra- and intersubject cerebral variability may have been a significant source of variance. This was assumed by us to be a condition for justifying the appropriateness of the study in the first place. The Superiority

of the Goldman-Rakic

Model

In this study, CFAs of the five prefrontal functional constructs drawn from human neuropsychology challenged several metainterpretations currently circulating in the literature. The initial CFA rejects any distinction between the constructs “planning (or the elaboration of strategy) and execution of sequences of planned responses” and “self-regulation of behavior in response to environmental contingencies (including one’s own errors).” The next CFA revealed that the measures postulated to load on the factor “maintenance of a nonautomatic cognitive or behavioral set” manifested poor convergent validity. Finally, CFAs testing the divergent validity of the prefrontal construct spatiotemporal segmentation and organization of events and of a nonprefrontal memory construct, were not supported. Despite the fact that the measures derived for the present study were not designed to test the Goldman-Rakic model, the CFA set up to test the model indicated good fit. Furthermore, the factorial validity of the Goldman-Rakic model was superior to two null models and two random models. These results suggest that prefrontal functional constructs proposed by Goldman-Rakic are likely to be more valid than those currently cited in the human neuropsychology literature. Analysis of Each of the Five Prefrontal Constructs Drawn from Human Neuropsychology Confronted by the Goldman-Rakic Model

The absence of divergent validity of the two factors Fl planning (or elaboration of strategy) and execution of sequences of planned responses measured by SOPT-E and MAZES-NPE and F2 self-regulation of behavior in response to environmental contingencies (including one’s own errors) measured by WCST-PE and MAZES-PE can be considered predictable within Goldman-Rakic’s model: (a) Sequencing tasks are considered by Goldman-Rakic to represent an extension of tasks involving delayed responding, in which it is the specific nature of the on-line representational memory (ex: verbal, visuospatial) that is critical. According to this concept, performance on the SOPT (visual and verbal) and on the Porteus Maze Test (visuospatial) should depend on distinct prefrontal contributions (thus the first factor should not cohere); (b) perseverative and nonperseverative errors on Porteus Mazes are both conceived as alterations of prefrontal visuospatial representational memory (thus MAZES-NPE and MAZES-PE would be predicted to fit onto the same factor); and (c) perseverative errors on the WCST (verbal repre-

OPPOSING

PREFRONTAL

FUNCTION

MODELS

63

sentational memory) are judged distinct from perseverative errors on the Porteus Maze Test (visuospatial representational memory) (thus the second factor should not cohere). The third factor, F3 maintenance of a nonautomatic cognitive or behavioral set, is not incompatible with Goldman-Rakic’s model. This function, translated into Goldman-Rakic’s terms, is behavioral control of prefrontal representational memory. However, the type of output involved in the task is not of negligible importance in Goldman-Rakic’s model. Thus, she would predict dissociation between the STROOP-E measure (verbal output) and the TRAILS-E and WCST-CB measures (manual output). Aside from these considerations, these three variables were weakly correlated. This may have been due to questionable criterionrelated validity of two of these, namely STROOP-E and TRAILS-E. Indeed, contrary to all the other dependent measures of this study for which systematic evidence of prefrontal sensitivity has been published, these two measures have only been declared sensitive, albeit by several authors. The fourth factor, F4 “spontaneity/sustained mental productivity” seems compatible with Goldman-Rakic’s model within which it would be termed a system of prefrontal self-activation. Of course, her model would predict some degree of dissociation on the basis of the distinct verbal and visuospatial representational memory involved in the two measures (VFLUENCY and D-FLUENCY). This position agrees with results from studies on human brain-damaged subjects. Indeed, right and left prefrontal lesions (dorsolateral and median) affect performance on these two tasks, the right-sided ones impairing D-FLUENCY more than the left, and the left-sided ones impairing V-FLUENCY more than the right (see Stuss & Benson, 1986, for a good review). The fifth factor, F5 spatiotemporal segmentation and organization of events, did not dissociate from a nonprefrontal memory construct in the present study. Studies of brain-damaged patients have shown double dissociation between recognition and recency judgments made on a similar set of stimuli (Milner, 1971; Ladavas, Umilta, & Provinciali, 1979). Temporohippocampal lesions impair recognition more than recency judgments and prefrontal lesions impair recency more than recognition judgments. For Goldman-Rakic, judgments of relative recency depend on the specific representational memory involved in the task. When recency judgments are made on events beyond short-term memory span, prefrontal representational memory becomes secondary to the threshold capacity of the hippocampal system. The results of our study support this latter point of view, but do not explain the double dissociation obtained from braindamaged patients. Several parameters are known to be involved in the performance of recency judgments. For example, the distance between the more recently judged stimulus and the probe may be the key element

64

DAIGNEAULT,

BRAtiN,

AND

WHITAKER

(possibility of an absolute recency judgment). In the case where this “distance” is short, the task may be infra-span, accessible to short-term memory, and realizeable despite temporohippocampal lesions. This issue requires carefully structured experimental investigation. With regard to the measure CONFUSION, it did not dissociate frontal from temporal patients in the study by Jetter and colleagues (1986) unless it was indexed to total recall, suggesting that the two measures (CONFUSION, VLEARNING) may not be independent. Our results challenge many associations, postulated in the neuropsychological literature, between several measures sensitive to prefrontal lesions and various “prefrontal” functions. They do not indicate, however, that the “prefrontal” functions proposed are false, nor that the postulated intertest associations are completely erroneous. Thus, one of the measures postulated as associated with a specific “prefrontal” function could be correctly associated and another not. Further Challenges by Goldman-Rakic to Interpretations of Prefrontal Brain Function Currently Held in Human Neuropsychology

In the human neuropsychological literature, some authors postulate that distractibility or alteration of directed attention, associated with prefrontal lesions, would be the consequence of alteration of a general and high order prefrontal function called “behavioral regulation by language” by Luria (1973)) “executive” by Stuss and Benson (1986), and “supervisory attentional system” by Shallice (1988). For Goldman-Rakic, these attentional deficits would be more circumscribed, being viewed as alterations of prefrontal representational memory which are specific to the task at hand. Thus, a focally lesioned patient could be distractible on one prefrontal task but not on another, drawing instead upon some other type of representational memory. Within the human neuropsychology literature, errors are categorized into prefrontal clusters (i.e., perseverative, automatistic, sequencing, etc.) regardless of the tasks on which they occur. Goldman-Rakic attaches less importance to these qualitative types of error patterns and assigns more importance to the types of tasks (verbal, visuospatial, etc.) on which the errors occur. 1. Perseverative errors. In human neuropsychology, these are often considered to reflect general behavioral or cognitive inflexibility, due to prefrontal dysfunction (Goldberg & Costa, 1986; Kolb & Wishaw, 1990; Mesulam, 1986; Milner, 1963). In fact, perseverative errors can occur massively in patients with posterorolandic lesions (Albert & Sandson, 1986; Allison, 1966; Buckingham, Whitaker, & Whitaker, 1981; Hecaen, Penfield, Bertrand, & Malmo, 1956; Lhermitte & Beauvois, 1973; Sandson & Albert, 1984; Yamadori, 1981). Goldman-Rakic’s position is compatible with the results of these researchers who have investigated verbal per-

OPPOSING PREFRONTAL

FUNCTION MODELS

65

severation in aphasics, and who have shown that perseverative errors result from alteration of specific functions involved in specific tasks (Allison & Hurwitz, 1967) or who have even shown that the degree of perseveration is a function more of specific functional impairment than the locus of the lesion (Albert & Sandson, 1986). This conception is further supported by results, from groups of patients with highly localized focal lesions, on the Design Fluency Task (Jones-Gotman & Milner, 1977). Thus, on the fixed condition of this task, impaired productivity by various groups of patients (right frontal, left frontal, right temporal, right parietal) is commensurate with their level of perseveration. 2. Errors associated with automatisms. These have occasionally been interpreted in the human neuropsychology literature as an impairment of “inhibitory” prefrontal function (Kolb & Whishaw, 1990; Malloy, 1987; Stuss & Benson, 1986). For Goldman-Rakic, this type of error in prefrontal patients is again a result of an alteration of the prefrontal representational memory required to accomplish a cognitive task. These errors cannot be assimilated, in her view, to “release” phenomena as suggested by emergence of primitive reflexes, since cognitive prefrontal circuits are always both inhibitory and excitatory. A careful consideration of prefrontal tasks eliciting errors resembling automatisms suggests that the required response is in fact a nonhabitual response which conflicts with an automatized stimulus-response association. 3. Sequencing errors. These are occasionally viewed, in the human neuropsychology literature, as a prefrontal “planning” deficit or as a difficulty of maintaining the order of planned responses (Milner, Petrides, & Smith, 1985; Petrides & Milner, 1982; Stuss & Benson, 1986). The ambiguity of the neuropsychological inference is due to the fact that sequencing tasks, such as the SOPT, are complex. However, Petrides (1991) recently interpreted the deficit associated with prefrontal lesions on this task as being due to alteration of a specific prefrontal working memory. This recent interpretation is closer to Goldman-Rakic’s. Thus, for Goldman-Rakic, sequencing represents a special case of delayed responding. By itself the planning of nonautomatic sequences of actions is, in her view, probably integrated at multiple levels of median, orbito, and dorsolateral prefrontal cortex and of posterorolandic cortex, working cooperatively. The specific dorsolateral prefrontal contribution would consist of on-line representational memory and regulation of preplanned nonautomatic action sequences. Goldman-Rakic’s model goes further than this; it would predict that patients with dorsolateral lesions could manifest spared planning on a sequencing task (ability to repeat the instructions correctly, explicitly formulating an appropriate strategy, if questioned) while still manifesting impaired performance on the task. In human neuropsychology, a general inhibitory function has occasionally been attributed to the orbital area of prefrontal cortex (Botez, 1987;

66

DAIGNEAULT,

BRAtiN,

AND

WHITAKER

Luria, 1973; Malloy, 1987; Stuss & Benson, 1986). However, GoldmanRakic argues that this interpretation is discordant with the physiological data currently available regarding output from the frontal lobes. It is possible that the association between inhibitory function and orbital cortex was a consequence of observer bias wherein “positive” behavior, more disruptive and more easily repertoried, has been disproportionately emphasized in these patients relative to “negative” behavior or inaction, which is nondisruptive and difficult to tally. Goldman-Rakic believes that orbital cortex may have privileged access to representations of reward and punishment. It is in this context then, rather than in a context of general disinhibition, that orbital cortical lesions could have severe repercussions on learned social contingencies. Other authors have also interpreted orbital function along similar lines (Grafman, 1988; Jouandet & Gazzaniga, 1979; Kolb & Wishaw, 1990; Mesulam, 1986). Conclusion Regarding Goldman-Rakic

Model

Goldman-Rakic’s model is well articulated and integrated. Many behavioral alterations known to result from prefrontal lesions in humans can be explained in interesting ways by this model. The model perhaps best explains how a bewildering array of cognitive, perceptual, and mnestic functions, which are spared following prefrontal lesions, are under-used in novel unstructured tasks. This effect of high novelty and low structure on otherwise spared abilities has been noted by many neuropsychologists (Botez, 1987; Damasio, 1985; Jouandet & Gazzaniga, 1979; Luria, 1973; Milner, 1963, 1964; Stuss & Benson, 1986; Walsh, 1978). The neuropsychological literature indicates a prefrontal hemispheric functional dominance, left-verbal, right-nonverbal, complementary with the lateral specialization of posterorolandic functions (Golden, Osmon, Moses, & Berg, 1981; McCarthy & Warrington, 1986; Moscovitch, 1979; Milner, 1971; Seron, 1978; Stuss & Benson, 1986). The antero-poster0 rolandic parallel and reported intraprefrontal dissociations (Milner, 1963; Stuss & Benson, 1986) support the idea of different representations accessed by different prefrontal areas. Such a prefrontal modular organization is proposed of course by Goldman-Rakic. In addition to its involvement in the simple regulation of new behavior in the absence of a structuring environment, Goldman-Rakic states that prefrontal representational memory could be involved in higher order cognition present only in humans (certain types of anticipation, problem solving, etc.). To further test Goldman-Rakic’s model, new clinical measures will have to be developed. These will have to be operationalized in such a manner as to measure simple types of prefrontal representational memory in accordance with specific connections of prefrontal sub-areas with extraprefrontal cortex, therefore involving different types of representations. These new tasks will have to be administered to patients with focal pre-

OPPOSING PREFRONTAL

FUNCTION MODELS

67

frontal lesions and should exhibit the following characteristics: (a) novelty-they should minimally depend on prior learning; (b) the discriminative stimulus should be removed prior to the response; (c) the discriminative stimulus should be very simple so as to minimally draw on extraprefrontal contribution; (d) the discriminative stimulus should be unimodal (verbal, visuospatial, sensorimotor, prosodic. . .); (e) the hippocampal contribution should be controlled; (f) execution speed should not be a performance factor; and (g) the nature of the response should be controlled (ex: oculomotor, verbal or manual). Such a research paradigm could more definitively validate or disconfirm this model. If this research program were to turn out supportive of the model, a second stage of research could be envisioned, investigating the effects of prefrontal lesions on more complex cognitive operations. The results of the present study should not be construed as more than a strong motivation to seriously consider Goldman-Rakic’s model for the human. They are not a definitive demonstration of the validity of the model. Good fit does not, of itself, guarantee construct validity of any model. Goldman-Rakic’s model is seemingly trapped in a kind of circularity wherein prefrontal and extraprefrontal functions are everywhere playing hide-and-seek as in a mirror arcade, not letting us clearly know whether they are operating inside or outside prefrontal cortex. Furthermore, a priori labeling of a functional construct can be exact, superficial or downright misguided. Consider for example, how easy it would be to label Fl “problem solving ability”, F2 “verbal ability”, F3 “visuospatial ability” and F4 “memory.” It is also worthy of mention that any multivariate psychometric investigation of existing standardized prefrontal measures will, at least in the short term, be limited by the generally low level of divergent validity of these measures. Indeed, prefrontal tests are “barely” prefrontal, but this is the best we have to work with at present. This well known characteristic of existing prefrontal measures, in addition to the present results indicating low validity of commonly invoked underlying constructs, advocates great prudence in the clinical interpretation of these tests. Also, it must be recognized that not all existing prefrontal tests were included in the present investigation. Had such prefrontal tests as the Verbal Concept Attainment Test, Shallice’s Cognitive Estimation Test, Petrides’s Conditional Associative Learning Tasks been included, the overall result might have been different. Finally, the present study does not remotely contribute to any sort of demonstration of criterion-related validity. This type of validity can only be demonstrated using brain-damaged patients. ACKNOWLEDGMENTS Fran&e Lussier, Anne DCcarie, and Brigitte Gilbert participated in subject recruitment and testing. Our gratitude is expressed to Dr. M. Petrides for putting his test, the SOPT,

68

DAIGNEAULT,

BRAUN, AND WHITAKER

at our disposal and to the granting agencies le Fonds pour la Formation de Chercheurs et I’Aide a la Recherche (Quebec), le ConseiI Quebecois de la Recherche Sociale, and la Fondation Lucie-Bruneau for their financial support. Y. Joanette and H. Cohen made useful constructive comments on early drafts and F. Richer led us to the relevant work of P. Goldman-Rakic.

REFERENCES Albert, M. L., & Sandson, J. 1986. Perseveration in aphasia. Cortex, 22, 103-115. Allison, R. 1966. Perseveration as a sign of diffuse and focal brain damage-II. British Medical Journal, 2, 1095-1101. Allison, R., & Hurwitz, L. 1967. On perseveration in aphasics. Brain, 90, 429-448. Bentler, P. M., & Bonett, D. G. 1980. Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88, 588-W. Benton, A. L., & Hamsher, K. de S. 1983. Multilingual Aphasia Examination: Manual of Instructions. Iowa: AJA Associates, Inc. Botez, M. I. 1987. Les syndromes du lobe frontal [Frontal lobe syndromes]. In M. I. Botez (Ed.), Neuropsychologie clinique et neurologie du comportement. Montreal: Les Presses du l’Universit6 de Montreal. Pp. 117-134. Brown, J. W. 1988. The life of the mind. Hillsdale: Lawrence Erlbaum Associates. Buckingham, H., Whitaker, H., & Whitaker H. A. 1981. On linguistic perseveration. In H. Whitaker & H. A. Whitaker (Eds.), Studies in neurolinguistics. New York: Academic Press. Vol. 4, pp. 329-352. Cole, D. A. 1987. Utility of confirmatory factor analysis in test validation research. Journal of Consulting and Clinical Psychology, 55, 584-594. Corsi, P. M. 1972. Human memory and the medical temporal region of the brain. Unpublished doctoral dissertation, McGill University, Montreal. Daigneault, S., Braiin, C. M. J., Proulx, R., & Gilbert, B. 1988. Quatre parametres influeqant la memoire de recence [Four parameters influencing recency judgments]. Resume des Communications du lleme Congres de la Societe Qutbecoise de Recherche en Psychologie, 71. [abstract] Damasio, A. R. 1985. The frontal lobes. In K. M. Heilman & E. Valenstein (Eds.), Clinical Neuropsychology. New York: Oxford Press. Pp. 339-375. Delis, D. C., Freeland, J., Kramer, J. H., & Kaplan, E. 1988. Integration clinical assessment cognitive neuroscience: Construct validation of the California Verbal Learning Test. Journal of Consulting and Clinical Psychology, 56, 123-130. Dillon, W. R., & Goldstein, M. 1984. Multivariate analysis: Methods and applications. New York: Wiley. Ellis, A. W., & Young, A. W. 1988. Human cognitive neuropsychology. Hillsdale: Lawrence Earlbaum Associates. Goldberg, E., & Costa, L. D. 1986. Qualitative indices in neuropsychological assessment: An extension of Luria’s approach to executive deficit following prefrontal lesions. In I. Grant & K. M. Adams (Eds.), Neuropsychological assessmentof neuropsychiatric disorders. New York: Oxford University Press. Pp. 48-64. Golden, C. J. 1978. Stroop Color nnd Word Test. Chicago: Stoelting Company. Golden, C. J., Osmon, D. C., Moses, Jr., J. A., & Berg, R. A. 1981. Interpretation of the Halstead-Reitan neuropsychological test battery: A casebook approach. New York: Grune & Stratton. Pp. 45-66. Goldman-Rakic, P. S. 1987. Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In F. Plum (Ed.), Handbook of Physiology: The Nervous System. New York: Oxford University Press. Vol. 5, pp. 373-417. Grafman, J. 1988. Plans, actions and mental sets: Managerial knowledge units in the frontal

OPPOSING PREFRONTAL

FUNCTION MODELS

69

lobes. In E. Perecman (Ed.), Integrations theory and practice in neuropsychology. Hillsdale: Lawrence Erlbaum Associates. Heaton, R. K. 1981. Wisconsin Card Sorting Test. Odessa: Psychological Assessment Recources Inc. HCcaen, H., Penfield, W., Bertrand, C., & Malmo, R. 1956.The syndromes of apractognosia due to lesions of the minor hemisphere. Archives of Neurology and Psychiatry, 75, 4of-434. Hecaen, H., & Ruel, J. 1981. Sieges lesionnels intrafrontaux et deficit au test de “fluence verbale” [Verbal fluency and sites of frontal lesions]. Revue de Neurologie, W7, 277284. Holst, P., & Vilkki, J. 1988. Effect of frontomedial lesions on performance on the Stroop Test and Word Fluency Tasks. Journal of Clinical and Experimental Neuropsychology, 10, 79. Jetter, W., Poser, U., Freeman, R. B., & Markowitsch, H. J. 1986. A verbal long term memory deficit in frontal lobe damaged patients. Cortex, 22, 229-242. Jones-Gotman, M., & Milner, B. 1977. Design fluency: The invention of nonsense drawing after focal cortical lesions. Neuropsychologia, 15, 653-674. Jones-Gotman, M., Ptito, A., & Zatorre, R. J. 1984. Deficits cognitifs associes aux lesions ctrebrales localisees. [Cognitive deficits asspciated with focal cerebral lesions]. Revue Quebecoise de Psychologie,

5, 83-104.

Joreskog, K. G., & Sorbom, D. 1984. Lisrel VI-Analysis by Maximum

Likelihood,

Instrumental

of Linear Structural Relationship Variables, and Least Squares Methods. Sweden:

Scientific Software, Inc. (3rd ed.) Jouandet, M., & Gazzaniga, M. S. 1979. The frontal lobes. In M. S. Gazzaniga (Ed.), Handbook of behavioral neurobiology. New York: Plenum Press. Vol. 2, pp. 25-59. Kolb, B., & Wishaw, I. Q. 1990. Fundamentals of Human Neuropsychology. New York: Freeman and Company. 3rd. ed., pp. 463-500. Ladavas, E., Umilta, C., & Provinciali, L. 1979. Hemisphere dependent cognitive performances in epileptic patients. Epilepsia, 20, 493-502. Lezak, M. D. 1983. Neuropsychological assessment. New York: Oxford Univ. Press. 2nd ed. Lhermitte, E., & Beauvois, M. F. 1973. A visual-speech disconnexion syndrome-report of a case with optic aphasia, agnostic alexia and colour agnosia. Brain, 96, 695-714. Luria, A. R. 1973. The Working Brain. New York: Basic Books, Inc. Luria, A. R. 1976. The neuropsychology of memory. Washington: Halstead Press. Luria, A. R. 1980. Disturbances of higher cortical functions with lesions of the frontal region. In A. R. Luria (Ed.), Higher cortical functions in man. New York: Basic Books Inc. (2nd ed.) Malloy, P. 1987. Frontal lobe dysfunction in obsessive-compulsive disorder. In E. Perecman (Ed.), The frontal lobe revisited. New York: The IRBN Press. McCarthy, R. A., & Warrington, E. K. 1986. Cognitive neuropsychology: A clinical introduction. Toronto: Academic Press. Pp. 343-364. Mesulam, M. M. 1986. Frontal cortex and behavior. Annals of Neurology, 19, 320-325. Miceli, G., Caltagirone, C., Gainotti, G., Masullo, C., & Silveri, M. C. 1981. Neuropsychological correlates of localized cerebral lesions in non-aphasic brain-damaged patients. Journal of Clinical Neuropsychology,

3, 53-63.

Mimer, B. 1963. Effects of different brain lesions on card sorting. Archives of Neurology, 9, 90-100. Mimer, B. 1964. Some effects of frontal lobectomy in man. In J. W. Warren & K. Akert (Eds.), The frontal granular cortex and behavior. New York: McGraw-Hill Book. Pp. 313-324.

70

DAIGNEAULT,

BRAUN, AND WHITAKER

Milner, B. 1971. Interhemispheric differences in the localization of psychological processes in man. British Medical Bulletin, 27, 272-277. Milner, B., & Petrides, M. 1984. Behavioural effects of frontal-lobe lesions in man. Trends in Neurosciences,

7, 403-407.

Milner, B., Petrides, M., & Smith, M. L. 1985. Frontal lobes and the temporal organisations of memory. Human Neurobiology, 4, 137-142. Moehle, K. B., Rasmussen, J. L., & Fitzhugh-Bell, K. B. 1987. Psychometric confirmation of neuropsychological theory. In J. M. Williams & C. J. Long (Eds.), Cognitive approaches to neuropsychology. New York: Plenum Press. Moscovitch, M. 1979. Information processing and the cerebral hemispheres. In M. S. Gazzaniga (Ed.), Handbook of behavioral neurobiology. New York: Plenum Press. Vol. 2, pp. 399-445. Newby, R. F., Hallenbeck, C. E., & Embretson, S. 1983. Confirmatory factor analysis of four neuropsychological models with a modified Halstead-Reitan Battery. Journal of Clinical Neuropsychology, 5, 115-133. Petrides, M. 1991, March. Functional specialization within the primate frontal cortex. In F. Benson (Chair), Basic Issues. Symposium conducted at the second annual conference of the Rotman Research Institute of Baycrest Center, Toronto. Petrides, M., & Milner, B. 1982. Deficits on subject-ordered tasks after frontal- and temporal-lobe lesions in man. Neuropsychologia, 28, 249-262. Petrie, A. 1949. Preliminary report of changes after prefrontal leucotomy. Journal of Mental Science, 95, 449-455. 1965. Porteus Maze Test: fifty years application.

Porteus, S. D. Porteus, S. D., lobotomy. Porteus, S. D.,

Palo Alto: Pacific Books. & De Monbrun Kepner, R. 1944. Mental changes after bilateral prefrontal Genetic Psychology Monographs, 29, 3-115. & Peters, H. N. 1947. Psychosurgery and test validity. Journal of Abnormal

and Social Psychology,

42, 473-475.

Ramier, A. M., & Hecaen, H. 1970. Role respectif des atteintes frontales et de la lateralisation lesionnelle dans les deficits de la “fluence verbale” [Effects of frontal lesions and lesion lateralization on defects of “Verbal Fluency”]. Revue Neurologique, W, 17-22. Reitan, R. M. 1958. Validity of the Trail Making Test as an indication of organic brain damage. Perpetual and Motor Skills, 8, 271-276. Rey, A. 1970. L’examen clinique en psychologie [The clinical examination in psychology]. Paris: Presses Universitaries de France. Russell, E. W. 1972. WAIS factor analysis with brain-damaged subjects using criterion measures. Journal of Consulting and Clinical Psychology, 39, 133-139. Sandson, J., & Albert, M. L. 1984. Varieties of perseveration. Neuropsychologia, 22, 715732. Schacter, D. L. 1987. Memory, amnesia, and frontal lobe dysfunction. Psychology, 15, 2136. Seron, X. 1978. Analyse neuropsychologique des lesions prefrontales chez l’homme [Neuropsychological analysis of human frontal lesions]. L’Annee Psychologique, 78, 183202. Shallice, T. 1988. The allocation of processing resources: Higher level control. IN T. Shallice (Ed.), From neuropsychology to mental structure. New York: Cambridge University Press. Pp. 328-352. Stuss, D. T., & Benson, D. F. 1984. Neuropsychological studies of the frontal lobes. Psychological

Bulletin,

95, 3-28.

Stuss, D. T., & Benson, D. F. 1986. The frontal lobes. New York: Raven Press. Stuss, D. T., Benson, D. F., Kaplan, E. F., Weir, W. S., Naeser, M. A., Lieberman, I.,

OPPOSING PREFRONTAL

FUNCIION

MODELS

71

& Ferrill, D. 1983. The involvement of orbitofrontal cerebrum in cognitive tasks. Neuropsychologia,

21, 235-248.

Walsh, K. W. 1978. The frontal lobes. In Neuropsychology: A Clinical Approach. New York: Livingston Churchill. Pp. 109-152. Wechsler, D. 1981. Wechsler adult intelligence scale-revised. New York: The Psychological Corporation. Wechsler, D. 1987. Wechsler memory scale-revised. New York: The Psychological Corporation. Wiegersma, S., Van der Scheer, E., & Human, R. 1990. Subjective ordering, short-term memory, and the frontal lobes. Neuropsychologia, 28, 95-98. Yamadori, A. 1981. Verbal perseveration in aphasia. Neuropsychologia, 19, 591-594.