- Email: [email protected]
Anisometry of Space Representation in Unilateral Neglect: Empirical Test of a Former Hypothesis
Consciousness and Cognition 8, 577–584 (1999) Article ID ccog.1999.0416, available online at http://www.idealibrary.com on
BRIEF REPORT Anisometry of Space Representation in Unilateral Neglect: Empirical Test of a Former Hypothesis Edoardo Bisiach, Marco Neppi-Mo`dona, and Rosanna Genero Dipartimento di Psicologia, Universita` di Torino, Turin, Italy
and Riccardo Pepi Casa di Cura Giovanni XXIII, Pianezza, Italy When left-neglect patients are required to extend horizontal segments to double their original length, relative left overextension is frequently observed. Less frequently, relative left underextension may also be found. It was hypothesized that this contrast could depend on the degree of horizontal anisometry of the medium for the representation of spatial properties. The present paper reports an experiment conducted in order to test that hypothesis, on the basis of which left overextension should be larger with shorter than with longer segments and with segments lying in the right rather than in the left hemispace. Although supportive, the results unveiled unexpected complications: the expected effect of line length was found only in neglect patients with frontal damage, while the expected effect of side of presentation was found only in neglect patients without frontal damage. 1999 Academic Press
INTRODUCTION
In a recent extensive investigation (Bisiach, Ricci, & Neppi Mo`dona, 1998) it was found that, when left-neglect patients are required to extend on either side horizontal segments to double their original length, relative left overextension is most frequently found. Although less frequently, relative left underextension may also be observed. The qualifier ‘‘relative’’ is added in order to underline the fact that left over- or underextension is independent of the absolute errors made by patients on the line extension task. Relative left over- or underextension, in fact, may be found when patients overextend line segments both left- and rightward, as well as when they underextend them in both of these directions. Relative left overextension was interpreted as the consequence of progressive contralesional relaxation of the medium for space representation, such that horizontal lines lying side by side are perceived as equal if that on the contralesional side is The authors thank William P. Banks and Bruce Bridgeman for extremely helpful suggestions and the administrator of Casa di Cura Papa Giovanni XXIII, Pietro Camerlengo, for his kind assistance. The research was supported by CNR and MURST grants to the first author. Address correspondence and reprint requests to Edoardo Bisiach, 22070 Lurago Marinone, (Como), Italy. Fax: ⫹⫹ 39 031 935097. 577 1053-8100/99 $30.00 Copyright 1999 by Academic Press All rights of reproduction in any form reserved.
578
BISIACH ET AL.
FIG. 1. (a) A lower degree of contralesional relaxation of the medium for space representation causing rightward bisection error (above) and left overextension (below, dashed line). (b) A higher degree of contralesional relaxation of the medium for space representation causing rightward bisection error and left underextension. The shaded surface represents the neglect area in which the overrelaxed medium cannot sustain conscious representation. (Modified from Bisiach et al., 1998.)
actually longer. Relative left underextension was interpreted as the consequence of contralesional overrelaxation of the representational medium preventing, beyond a critical point, conscious representation (Fig. 1). Partial support for that twofold hypothesis was the higher severity of neglect (as assessed through line bisection and H cancellation) in patients who showed relative left underextension. On that hypothesis, a higher degree of relative left overextension would, on the other hand, be expected (1) with shorter lines and (2) with lines located on the ipsilesional side: in both cases, indeed, more room would be left (on the contralesional side of space) before the point is reached beyond which conscious representation is
SPACE REPRESENTATION IN UNILATERAL NEGLECT
579
abated. The experiment reported in this paper was conducted in order to test these two predictions. METHOD
Subjects Fifteen right brain-damaged patients with neglect (N⫹),1 15 right brain-damaged patients without neglect (N⫺), and 10 subjects without past or present brain damage (C) gave informed consent to participate in the experiment. All patients and control subjects were righthanders by self-report. The presence and location of brain lesions were assessed by means of computerized tomography (CT). Inclusion in the N⫹ group was based on a difference of five or more omissions between the left and right halves of the stimulus array on a letter H cancellation task (a slight variant of Diller & Weinberg, 1977) and/or a mean rightward error of 5 mm or more on five bisection trials with horizontal black lines 150 mm long and 1 mm thick, centred on a horizontal A4 sheet of white paper. In the analysis of results, N⫹ patients were further subdivided into two subgroups, N⫹F⫹ and N⫹F⫺, according to the presence/absence of frontal lobe damage (cf. Table 1). The N⫹F⫹ subgroup included 8 patients (Nos. 6, 8–14); the N⫹F⫺ subgroup included 7 patients (Nos.; 1–5, 7, 15). The three main experimental groups did not significantly differ with respect to age. By contrast, there was a significant intergroup difference with respect to educational level [F(2, 37) ⫽ 3.517; p ⫽ .040]. Posthoc analysis showed that the mean educational level was lower in group N⫹ (5.00 years; SD ⫽ 2.65) than in group N⫺(7.20 years; SD ⫽ 2.51) [t (28) ⫽ 2.335; p ⫽ .041]. Stimuli and Procedure Stimuli were computer-generated 1.50-mm-thick black horizontal lines lying centrally on a thinner (0.38-mm) black line drawn along the whole longer meridian of A4 sheets of white paper. Subjects were given a randomized sequence of 80 trials on which they had to draw a leftward (LE) or rightward (RE) extension doubling the original length of 10-mm lines (40 trials; 20 LE and 20 RE) and 60-mm lines (40 trials; 20 LE and 20 RE). Extensions were drawn along the thinner line underlying the stimulus, in order to prevent upward or downward deflections. The task was executed on either side of egocentric space: on 40 trials (10 for each line length and side of extension) the right and, in the remaining 40 trials, the left border of the sheet of paper on which the stimulus was printed was placed (flat on a desk) coincident with the forward extension of the midsagittal plane of the subject’s trunk, at a comfortable writing distance.
1 Abbreviations used: C, control subjects; LE, left extension; LE/RE, ratio of left/right extension; N⫹, neglect patients; N⫹F⫹, neglect patients with lesions confined or extending to the frontal lobe; N⫹F⫺, neglect patients with lesions sparing the frontal lobe; N⫺, right brain-damaged patients without neglect; RE, right extension.
580
BISIACH ET AL.
TABLE 1 Demographic, Clinical, and Experimental Data of N⫹ Patients Length of illness (days)
Lesion (CT scan)
Etiology
H Cancellation
Bisection 12.50 (5.74) 25.45 (8.73) 2.10 (3.13) 18.30 (5.85) 2.85 (6.06) 8.60 (3.39) 17.90 (23.85) 15.55 (4.73) ⫺1.85 (4.15) 3.50 (3.83) ⫺6.70 (5.03) 5.60 (7.24) 26.70 (11.22) 8.80 (3.32) 26.60 (5.32)
Sex
Age
Education (Years)
1
F
74
10
295
TP
I
143
2
F
72
5
147
TP
I
276
3
F
65
0
410
TP
I
43
4
F
61
4
29
O bg
I
247
5
M
66
5
21
TP ins
I
244
6
F
72
5
161
FT bg ci
I
8
7
F
74
0
392
P
I
0
8
F
66
5
313
FTO
I
202
9
F
67
6
68
F
H
52
10
F
51
7
259
FP
I
37
11
F
55
5
146
FT bg
I
49
12
F
66
5
268
FPT
I
161
13
M
66
5
108
FTO
I
122
14
M
71
9
1008
FT ci
I
21
15
M
77
4
55
TP
I
61
Patient
Note. Abbreviations used: bg, basal ganglia; F, frontal; O, occipital; P, parietal; T, temporal; H, hemorrhage; I, infarction; cancellation score (see Method); bisection, mean rightward or leftward (⫺) error (mm) and SD.
RESULTS
Table 1 reports demographic and clinical data of N⫹ patients, as well as their performance on preliminary tests: H cancellation and bisection of five 150-mm horizontal lines. The H cancellation score is the number of letters (targets and nontargets) lying to the left of a border connecting the leftmost H canceled by the patient on the first line of letters with the leftmost H canceled on the second line and so on. The reason for using such a score as a rough estimate of the extent of the neglected area is given elsewhere (Bisiach, Ricci, Lai, De Tanti & Inzaghi, 1999). Mean LE, RE, and LE/RE (and SDs) of groups N⫹ (further subdivided into N⫹F⫹ and N⫹F⫺), N⫺, and C are reported in Tables 2 and 3. A cumulative ANOVA on LE/RE could not be performed because unequal
581
SPACE REPRESENTATION IN UNILATERAL NEGLECT
TABLE 2 Mean Extension (mm) and SDs of Groups N⫹ (N⫹F⫹, N⫹F⫺), N⫺, and C
N⫹ N⫹F⫹ N⫹F⫺ N⫺ C
L-E 1L
R-E 1L
L-E 1R
R-E 1R
L-E 6L
R-E 6L
L-E 6R
R-E 6R
20.840 (6.780) 19.150 (6.156) 22.771 (7.451) 11.333 (1.309) 11.320 (1.160)
18.173 (6.424) 16.188 (3.981) 20.443 (8.157) 11.453 (1.282) 11.460 (.712)
21.573 (8.133) 19.563 (5.398) 23.871 (10.432) 11.647 (1.427) 11.220 (.928)
16.887 (4.284) 16.688 (5.548) 17.114 (2.607) 11.473 (1.090) 11.780 (1.003)
58.020 (12.200) 53.575 (14.521) 63.100 (6.717) 59.047 (3.372) 59.510 (2.427)
53.160 (9.390) 52.038 (11.001) 54.443 (7.804) 59.820 (2.716) 61.230 (1.544)
65.350 (14.170) 54.400 (8.131) 77.871 (6.980) 59.373 (3.052) 59.640 (2.501)
51.447 (8.713) 51.313 (7.777) 51.600 (10.322) 59.787 (2.981) 61.310 (1.746)
Note. Abbreviations used: L-E, left extension; R-E, right extension; 1L, 1-cm line in left hemispace; 1R, 1-cm line in right hemispace; 6L, 6-cm line in left hemispace; 6R, 6-cm line in right hemispace.
variances were found on Levene’s test (Stevens, 1990) across the experimental groups. On two separate repeated-measures ANOVAs on LE/RE in groups C and N⫺ the effects of line length and hemispace, as well as their interaction, were not significant. Two separate repeated-measures ANOVAs on LE/RE were also carried out with subgroups N⫹F⫹ and N⫹F⫺ because unequal variances found on Levene’s test between these subgroups of patients prevented a unitary analysis. In subgroup N⫹F⫹ the effect of line length was significant [F(1, 7) ⫽ 45.96; p ⬍ .0001]; the effect of hemispace and the interaction between line length and hemispace were not significant. In subgroup N⫹F⫺, by contrast, the effect of hemispace was significant [F(1, 6) ⫽ 6.24; p ⫽ .047], while the effect of line length and the interaction between line length and hemispace were not significant. As shown in Table 3 and Fig. 2, LE/RE increased in subgroup N⫹F⫹ from long to short lines, irrespective of the side of presentation. It increased in subgroup N⫹F⫺ from the left to the right side of presentation, irrespective of line length. TABLE 3 Mean L-E/R-E of Groups N⫹ (N⫹F⫹, N⫹F⫺), N⫺, and C
N⫹ N⫹F⫹ N⫹F⫺ N⫺ C
L-E/R-E 1L
L-E/R-E 1R
L-E/R-E 6L
L-E/R-E 6R
1.165 (.225) 1.176 (.178) 1.152 (.284) .991 (.050) .988 (.082)
1.274 (.305) 1.201 (.180) 1.358 (.406) 1.014 (.049) .953 (.043)
1.100 (.200) 1.027 (.165) 1.181 (.216) .988 (.054) .972 (.039)
1.305 (.392) 1.063 (.081) 1.581 (.429) .994 (.042) .973 (.030)
Note. Abbreviations used: L-E/R-E, ratio left extension/right extension; 1L, 1-cm line in left hemispace; 1R, 1-cm line in right hemispace; 6L, 6-cm line in left hemispace; 6R, 6-cm line in right hemispace. SDs within parentheses.
582
BISIACH ET AL.
FIG. 2. Effects of line length and side of presentation on relative left overextension by neglect patients on the line extension task.
Given the small sizes of subgroups N⫹F⫹ and N⫹F⫺, the lack of significance of some effects and interactions could be due to insufficient power of the statistical analysis. A retrospective power analysis (with the α level set at 0.05) showed that this was indeed the case, in subgroup N⫹F⫺, with the effect of line length (power ⫽ 0.133; effect size ⫽ 0.139) and the interaction between line length and hemispace (power ⫽ 0.15; effect size ⫽ 0.163). From inspection of power tables it can be inferred that increasing subgroup N⫹F⫺ to 25 patients with regard to the effect of line length, or 20 patients with regard to the interaction between line length and hemispace, would have been sufficient to raise the statistical power to 0.8, which is considered an acceptable level (Cohen, 1988). In other words, increasing the size of subgroup N⫹F⫺ would have been likely to lead to the rejection of the null hypothesis, with regard to both the effect of line length and the interaction between line length and hemispace. Correlations between LE/RE and the scores of line bisection and H cancellation were not significant, with the only exception being a strong negative correlation, in subgroup N⫹F⫹, between LE/RE with short segments in the right hemispace and line-bisection scores [r (6) ⫽ ⫺0.83; p ⫽ .010]. This datum, however, cannot be accepted without reservations, considering the high number of possible correlations between line-bisection scores and LE/RE values. DISCUSSION
As a whole, the results of the experiment confirm the two predictions stemming from the hypothesis that, in left neglect, relative left underextension on line extension tasks is due to contralesional overrelaxation of the medium for space representation: altogether, LE/RE was indeed found to increase (1) from long to short segments and (2), in agreement with findings by Nico, Galati, and Incoccia (1999), from the left to the right side of presentation. None the less, the first prediction found confirmation in subgroup N⫹F⫹ but not in subgroup N⫹F⫺, whereas the second found confirmation in subgroup N⫹F⫺ but not in subgroup N⫹F⫹.
SPACE REPRESENTATION IN UNILATERAL NEGLECT
583
This remarkable discrepancy could be illusory and only due to an insufficient number of observations. However, while a slight tendency of LE/RE to increase from the left to the right side was also evident in subgroup N⫹F⫹ (cf. Table 3 and Fig. 2), the increase of LE/RE from long to short segments found in subgroup N⫹F⫹ is in sharp contrast to the opposite tendency evident in subgroup N⫹F⫺, a tendency corroborated by the power analysis. A tentative explanation of this puzzling finding can, however, be considered. On the one hand, the absent (or marginal) effect of side of presentation in subgroup N⫹F⫹ could be due to the possibility that neglect affecting patients of this group was prevailingly framed in allocentric rather than in egocentric coordinates (Bisiach, 1997). In this case, the patients’ error on the line extension task would in fact be completely (or largely) independent of the location of the stimulus array with respect to their trunk’s sagittal midplane. On the other hand, the critical point to the left of which the progressive relaxation of the representational medium is no longer able to sustain consciousness could have been more peripheral in patients of subgroup N⫹F⫺ than in those of subgroup N⫹F⫹. Figure 1b shows that in this case N⫹F⫺ patients would have benefitted from a larger representational extent within the limits of which LE/RE would in fact increase, rather than decrease, from short to long segments. Suppose the (shadowed) neglect area were located in Fig. 1b to the left of the leftmost vertical line of the diagram and consider, for example, two segments immediately located to the left of the rightmost vertical line: one shorter (1 unit) and one longer (3 units); the overshoot made in extending each segment leftward to double the original length (i.e., to 2 and 6 units, respectively) would be proportionally larger with the longer than with the shorter segment. It can easily be inferred that in the case of rightward extension the undershoot would also be proportionally larger with the longer than with the shorter segment. LE/RE would therefore increase, as actually found in subgroup N⫹F⫺, from short to long segments. A further complication is contributed by the strong negative correlation found only between the size of the rightward error on line bisection and LE/RE in subgroup N⫹F⫹, with short segments in the right hemispace. Though in agreement with earlier findings (Bisiach et al., 1998) and with the hypothesis that relative overextension on a line extension task could reverse to relative underextension in patients with severe left neglect, such a result is itself suggestive of a state of affairs more complex then initially envisaged. Further investigation should try not only to answer the questions we have here discussed and to understand why line extension is differently affected by lesion involving different brain structures, it should also address apparently minor points, such as the tendency, specific to neglect patients (cf. Table 2), to exaggeratedly overextend short (but not long) segments on both sides. REFERENCES Bisiach, E. (1997). The spatial features of unilateral neglect. In P. Thier & H.-O. Karnath (Eds.), Parietal lobe contributions to orientation in 3D space (pp. 465–495). Heidelberg: Springer-Verlag. Bisiach, E., Ricci, R., Lai, E., De Tanti, A., & Inzaghi, M. G. (1999). Unilateral neglect and disambiguation of the Necker cube. Brain, 122, 131–140.
584
BISIACH ET AL.
Bisiach, E., Ricci, R., & Neppi Mo`dona, M. (1998). Visual awareness and anisometry of space representation in unilateral neglect: A panoramic investigation by means of a line extension task. Consciousness and Cognition, 7, 327–355. Cohen, J. (1988). Statistical power analysis for the behavioural sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Diller, L., & Weinberg, J. (1977). Hemi-inattention in rehabilitation: The evolution of a rational remediation program. In E.A. Weinstein & R.P. Friedland (Eds.), Hemi-inattention and hemisphere specialization (pp. 51–82). New York: Raven Press. Nico, D., Galati, G., & Incoccia, C. (1999). The endpoints’ task: An analysis of length reproduction in unilateral neglect. Neuropsychologia, 37, 1181–1188. Stevens, J.P. (1990). Intermediate statistics: Modern approach. Hillsdale, NJ: Erlbaum. Received June 24, 1999
BRIEF REPORT Anisometry of Space Representation in Unilateral Neglect: Empirical Test of a Former Hypothesis Edoardo Bisiach, Marco Neppi-Mo`dona, and Rosanna Genero Dipartimento di Psicologia, Universita` di Torino, Turin, Italy
and Riccardo Pepi Casa di Cura Giovanni XXIII, Pianezza, Italy When left-neglect patients are required to extend horizontal segments to double their original length, relative left overextension is frequently observed. Less frequently, relative left underextension may also be found. It was hypothesized that this contrast could depend on the degree of horizontal anisometry of the medium for the representation of spatial properties. The present paper reports an experiment conducted in order to test that hypothesis, on the basis of which left overextension should be larger with shorter than with longer segments and with segments lying in the right rather than in the left hemispace. Although supportive, the results unveiled unexpected complications: the expected effect of line length was found only in neglect patients with frontal damage, while the expected effect of side of presentation was found only in neglect patients without frontal damage. 1999 Academic Press
INTRODUCTION
In a recent extensive investigation (Bisiach, Ricci, & Neppi Mo`dona, 1998) it was found that, when left-neglect patients are required to extend on either side horizontal segments to double their original length, relative left overextension is most frequently found. Although less frequently, relative left underextension may also be observed. The qualifier ‘‘relative’’ is added in order to underline the fact that left over- or underextension is independent of the absolute errors made by patients on the line extension task. Relative left over- or underextension, in fact, may be found when patients overextend line segments both left- and rightward, as well as when they underextend them in both of these directions. Relative left overextension was interpreted as the consequence of progressive contralesional relaxation of the medium for space representation, such that horizontal lines lying side by side are perceived as equal if that on the contralesional side is The authors thank William P. Banks and Bruce Bridgeman for extremely helpful suggestions and the administrator of Casa di Cura Papa Giovanni XXIII, Pietro Camerlengo, for his kind assistance. The research was supported by CNR and MURST grants to the first author. Address correspondence and reprint requests to Edoardo Bisiach, 22070 Lurago Marinone, (Como), Italy. Fax: ⫹⫹ 39 031 935097. 577 1053-8100/99 $30.00 Copyright 1999 by Academic Press All rights of reproduction in any form reserved.
578
BISIACH ET AL.
FIG. 1. (a) A lower degree of contralesional relaxation of the medium for space representation causing rightward bisection error (above) and left overextension (below, dashed line). (b) A higher degree of contralesional relaxation of the medium for space representation causing rightward bisection error and left underextension. The shaded surface represents the neglect area in which the overrelaxed medium cannot sustain conscious representation. (Modified from Bisiach et al., 1998.)
actually longer. Relative left underextension was interpreted as the consequence of contralesional overrelaxation of the representational medium preventing, beyond a critical point, conscious representation (Fig. 1). Partial support for that twofold hypothesis was the higher severity of neglect (as assessed through line bisection and H cancellation) in patients who showed relative left underextension. On that hypothesis, a higher degree of relative left overextension would, on the other hand, be expected (1) with shorter lines and (2) with lines located on the ipsilesional side: in both cases, indeed, more room would be left (on the contralesional side of space) before the point is reached beyond which conscious representation is
SPACE REPRESENTATION IN UNILATERAL NEGLECT
579
abated. The experiment reported in this paper was conducted in order to test these two predictions. METHOD
Subjects Fifteen right brain-damaged patients with neglect (N⫹),1 15 right brain-damaged patients without neglect (N⫺), and 10 subjects without past or present brain damage (C) gave informed consent to participate in the experiment. All patients and control subjects were righthanders by self-report. The presence and location of brain lesions were assessed by means of computerized tomography (CT). Inclusion in the N⫹ group was based on a difference of five or more omissions between the left and right halves of the stimulus array on a letter H cancellation task (a slight variant of Diller & Weinberg, 1977) and/or a mean rightward error of 5 mm or more on five bisection trials with horizontal black lines 150 mm long and 1 mm thick, centred on a horizontal A4 sheet of white paper. In the analysis of results, N⫹ patients were further subdivided into two subgroups, N⫹F⫹ and N⫹F⫺, according to the presence/absence of frontal lobe damage (cf. Table 1). The N⫹F⫹ subgroup included 8 patients (Nos. 6, 8–14); the N⫹F⫺ subgroup included 7 patients (Nos.; 1–5, 7, 15). The three main experimental groups did not significantly differ with respect to age. By contrast, there was a significant intergroup difference with respect to educational level [F(2, 37) ⫽ 3.517; p ⫽ .040]. Posthoc analysis showed that the mean educational level was lower in group N⫹ (5.00 years; SD ⫽ 2.65) than in group N⫺(7.20 years; SD ⫽ 2.51) [t (28) ⫽ 2.335; p ⫽ .041]. Stimuli and Procedure Stimuli were computer-generated 1.50-mm-thick black horizontal lines lying centrally on a thinner (0.38-mm) black line drawn along the whole longer meridian of A4 sheets of white paper. Subjects were given a randomized sequence of 80 trials on which they had to draw a leftward (LE) or rightward (RE) extension doubling the original length of 10-mm lines (40 trials; 20 LE and 20 RE) and 60-mm lines (40 trials; 20 LE and 20 RE). Extensions were drawn along the thinner line underlying the stimulus, in order to prevent upward or downward deflections. The task was executed on either side of egocentric space: on 40 trials (10 for each line length and side of extension) the right and, in the remaining 40 trials, the left border of the sheet of paper on which the stimulus was printed was placed (flat on a desk) coincident with the forward extension of the midsagittal plane of the subject’s trunk, at a comfortable writing distance.
1 Abbreviations used: C, control subjects; LE, left extension; LE/RE, ratio of left/right extension; N⫹, neglect patients; N⫹F⫹, neglect patients with lesions confined or extending to the frontal lobe; N⫹F⫺, neglect patients with lesions sparing the frontal lobe; N⫺, right brain-damaged patients without neglect; RE, right extension.
580
BISIACH ET AL.
TABLE 1 Demographic, Clinical, and Experimental Data of N⫹ Patients Length of illness (days)
Lesion (CT scan)
Etiology
H Cancellation
Bisection 12.50 (5.74) 25.45 (8.73) 2.10 (3.13) 18.30 (5.85) 2.85 (6.06) 8.60 (3.39) 17.90 (23.85) 15.55 (4.73) ⫺1.85 (4.15) 3.50 (3.83) ⫺6.70 (5.03) 5.60 (7.24) 26.70 (11.22) 8.80 (3.32) 26.60 (5.32)
Sex
Age
Education (Years)
1
F
74
10
295
TP
I
143
2
F
72
5
147
TP
I
276
3
F
65
0
410
TP
I
43
4
F
61
4
29
O bg
I
247
5
M
66
5
21
TP ins
I
244
6
F
72
5
161
FT bg ci
I
8
7
F
74
0
392
P
I
0
8
F
66
5
313
FTO
I
202
9
F
67
6
68
F
H
52
10
F
51
7
259
FP
I
37
11
F
55
5
146
FT bg
I
49
12
F
66
5
268
FPT
I
161
13
M
66
5
108
FTO
I
122
14
M
71
9
1008
FT ci
I
21
15
M
77
4
55
TP
I
61
Patient
Note. Abbreviations used: bg, basal ganglia; F, frontal; O, occipital; P, parietal; T, temporal; H, hemorrhage; I, infarction; cancellation score (see Method); bisection, mean rightward or leftward (⫺) error (mm) and SD.
RESULTS
Table 1 reports demographic and clinical data of N⫹ patients, as well as their performance on preliminary tests: H cancellation and bisection of five 150-mm horizontal lines. The H cancellation score is the number of letters (targets and nontargets) lying to the left of a border connecting the leftmost H canceled by the patient on the first line of letters with the leftmost H canceled on the second line and so on. The reason for using such a score as a rough estimate of the extent of the neglected area is given elsewhere (Bisiach, Ricci, Lai, De Tanti & Inzaghi, 1999). Mean LE, RE, and LE/RE (and SDs) of groups N⫹ (further subdivided into N⫹F⫹ and N⫹F⫺), N⫺, and C are reported in Tables 2 and 3. A cumulative ANOVA on LE/RE could not be performed because unequal
581
SPACE REPRESENTATION IN UNILATERAL NEGLECT
TABLE 2 Mean Extension (mm) and SDs of Groups N⫹ (N⫹F⫹, N⫹F⫺), N⫺, and C
N⫹ N⫹F⫹ N⫹F⫺ N⫺ C
L-E 1L
R-E 1L
L-E 1R
R-E 1R
L-E 6L
R-E 6L
L-E 6R
R-E 6R
20.840 (6.780) 19.150 (6.156) 22.771 (7.451) 11.333 (1.309) 11.320 (1.160)
18.173 (6.424) 16.188 (3.981) 20.443 (8.157) 11.453 (1.282) 11.460 (.712)
21.573 (8.133) 19.563 (5.398) 23.871 (10.432) 11.647 (1.427) 11.220 (.928)
16.887 (4.284) 16.688 (5.548) 17.114 (2.607) 11.473 (1.090) 11.780 (1.003)
58.020 (12.200) 53.575 (14.521) 63.100 (6.717) 59.047 (3.372) 59.510 (2.427)
53.160 (9.390) 52.038 (11.001) 54.443 (7.804) 59.820 (2.716) 61.230 (1.544)
65.350 (14.170) 54.400 (8.131) 77.871 (6.980) 59.373 (3.052) 59.640 (2.501)
51.447 (8.713) 51.313 (7.777) 51.600 (10.322) 59.787 (2.981) 61.310 (1.746)
Note. Abbreviations used: L-E, left extension; R-E, right extension; 1L, 1-cm line in left hemispace; 1R, 1-cm line in right hemispace; 6L, 6-cm line in left hemispace; 6R, 6-cm line in right hemispace.
variances were found on Levene’s test (Stevens, 1990) across the experimental groups. On two separate repeated-measures ANOVAs on LE/RE in groups C and N⫺ the effects of line length and hemispace, as well as their interaction, were not significant. Two separate repeated-measures ANOVAs on LE/RE were also carried out with subgroups N⫹F⫹ and N⫹F⫺ because unequal variances found on Levene’s test between these subgroups of patients prevented a unitary analysis. In subgroup N⫹F⫹ the effect of line length was significant [F(1, 7) ⫽ 45.96; p ⬍ .0001]; the effect of hemispace and the interaction between line length and hemispace were not significant. In subgroup N⫹F⫺, by contrast, the effect of hemispace was significant [F(1, 6) ⫽ 6.24; p ⫽ .047], while the effect of line length and the interaction between line length and hemispace were not significant. As shown in Table 3 and Fig. 2, LE/RE increased in subgroup N⫹F⫹ from long to short lines, irrespective of the side of presentation. It increased in subgroup N⫹F⫺ from the left to the right side of presentation, irrespective of line length. TABLE 3 Mean L-E/R-E of Groups N⫹ (N⫹F⫹, N⫹F⫺), N⫺, and C
N⫹ N⫹F⫹ N⫹F⫺ N⫺ C
L-E/R-E 1L
L-E/R-E 1R
L-E/R-E 6L
L-E/R-E 6R
1.165 (.225) 1.176 (.178) 1.152 (.284) .991 (.050) .988 (.082)
1.274 (.305) 1.201 (.180) 1.358 (.406) 1.014 (.049) .953 (.043)
1.100 (.200) 1.027 (.165) 1.181 (.216) .988 (.054) .972 (.039)
1.305 (.392) 1.063 (.081) 1.581 (.429) .994 (.042) .973 (.030)
Note. Abbreviations used: L-E/R-E, ratio left extension/right extension; 1L, 1-cm line in left hemispace; 1R, 1-cm line in right hemispace; 6L, 6-cm line in left hemispace; 6R, 6-cm line in right hemispace. SDs within parentheses.
582
BISIACH ET AL.
FIG. 2. Effects of line length and side of presentation on relative left overextension by neglect patients on the line extension task.
Given the small sizes of subgroups N⫹F⫹ and N⫹F⫺, the lack of significance of some effects and interactions could be due to insufficient power of the statistical analysis. A retrospective power analysis (with the α level set at 0.05) showed that this was indeed the case, in subgroup N⫹F⫺, with the effect of line length (power ⫽ 0.133; effect size ⫽ 0.139) and the interaction between line length and hemispace (power ⫽ 0.15; effect size ⫽ 0.163). From inspection of power tables it can be inferred that increasing subgroup N⫹F⫺ to 25 patients with regard to the effect of line length, or 20 patients with regard to the interaction between line length and hemispace, would have been sufficient to raise the statistical power to 0.8, which is considered an acceptable level (Cohen, 1988). In other words, increasing the size of subgroup N⫹F⫺ would have been likely to lead to the rejection of the null hypothesis, with regard to both the effect of line length and the interaction between line length and hemispace. Correlations between LE/RE and the scores of line bisection and H cancellation were not significant, with the only exception being a strong negative correlation, in subgroup N⫹F⫹, between LE/RE with short segments in the right hemispace and line-bisection scores [r (6) ⫽ ⫺0.83; p ⫽ .010]. This datum, however, cannot be accepted without reservations, considering the high number of possible correlations between line-bisection scores and LE/RE values. DISCUSSION
As a whole, the results of the experiment confirm the two predictions stemming from the hypothesis that, in left neglect, relative left underextension on line extension tasks is due to contralesional overrelaxation of the medium for space representation: altogether, LE/RE was indeed found to increase (1) from long to short segments and (2), in agreement with findings by Nico, Galati, and Incoccia (1999), from the left to the right side of presentation. None the less, the first prediction found confirmation in subgroup N⫹F⫹ but not in subgroup N⫹F⫺, whereas the second found confirmation in subgroup N⫹F⫺ but not in subgroup N⫹F⫹.
SPACE REPRESENTATION IN UNILATERAL NEGLECT
583
This remarkable discrepancy could be illusory and only due to an insufficient number of observations. However, while a slight tendency of LE/RE to increase from the left to the right side was also evident in subgroup N⫹F⫹ (cf. Table 3 and Fig. 2), the increase of LE/RE from long to short segments found in subgroup N⫹F⫹ is in sharp contrast to the opposite tendency evident in subgroup N⫹F⫺, a tendency corroborated by the power analysis. A tentative explanation of this puzzling finding can, however, be considered. On the one hand, the absent (or marginal) effect of side of presentation in subgroup N⫹F⫹ could be due to the possibility that neglect affecting patients of this group was prevailingly framed in allocentric rather than in egocentric coordinates (Bisiach, 1997). In this case, the patients’ error on the line extension task would in fact be completely (or largely) independent of the location of the stimulus array with respect to their trunk’s sagittal midplane. On the other hand, the critical point to the left of which the progressive relaxation of the representational medium is no longer able to sustain consciousness could have been more peripheral in patients of subgroup N⫹F⫺ than in those of subgroup N⫹F⫹. Figure 1b shows that in this case N⫹F⫺ patients would have benefitted from a larger representational extent within the limits of which LE/RE would in fact increase, rather than decrease, from short to long segments. Suppose the (shadowed) neglect area were located in Fig. 1b to the left of the leftmost vertical line of the diagram and consider, for example, two segments immediately located to the left of the rightmost vertical line: one shorter (1 unit) and one longer (3 units); the overshoot made in extending each segment leftward to double the original length (i.e., to 2 and 6 units, respectively) would be proportionally larger with the longer than with the shorter segment. It can easily be inferred that in the case of rightward extension the undershoot would also be proportionally larger with the longer than with the shorter segment. LE/RE would therefore increase, as actually found in subgroup N⫹F⫺, from short to long segments. A further complication is contributed by the strong negative correlation found only between the size of the rightward error on line bisection and LE/RE in subgroup N⫹F⫹, with short segments in the right hemispace. Though in agreement with earlier findings (Bisiach et al., 1998) and with the hypothesis that relative overextension on a line extension task could reverse to relative underextension in patients with severe left neglect, such a result is itself suggestive of a state of affairs more complex then initially envisaged. Further investigation should try not only to answer the questions we have here discussed and to understand why line extension is differently affected by lesion involving different brain structures, it should also address apparently minor points, such as the tendency, specific to neglect patients (cf. Table 2), to exaggeratedly overextend short (but not long) segments on both sides. REFERENCES Bisiach, E. (1997). The spatial features of unilateral neglect. In P. Thier & H.-O. Karnath (Eds.), Parietal lobe contributions to orientation in 3D space (pp. 465–495). Heidelberg: Springer-Verlag. Bisiach, E., Ricci, R., Lai, E., De Tanti, A., & Inzaghi, M. G. (1999). Unilateral neglect and disambiguation of the Necker cube. Brain, 122, 131–140.
584
BISIACH ET AL.
Bisiach, E., Ricci, R., & Neppi Mo`dona, M. (1998). Visual awareness and anisometry of space representation in unilateral neglect: A panoramic investigation by means of a line extension task. Consciousness and Cognition, 7, 327–355. Cohen, J. (1988). Statistical power analysis for the behavioural sciences (2nd ed.). Hillsdale, NJ: Erlbaum. Diller, L., & Weinberg, J. (1977). Hemi-inattention in rehabilitation: The evolution of a rational remediation program. In E.A. Weinstein & R.P. Friedland (Eds.), Hemi-inattention and hemisphere specialization (pp. 51–82). New York: Raven Press. Nico, D., Galati, G., & Incoccia, C. (1999). The endpoints’ task: An analysis of length reproduction in unilateral neglect. Neuropsychologia, 37, 1181–1188. Stevens, J.P. (1990). Intermediate statistics: Modern approach. Hillsdale, NJ: Erlbaum. Received June 24, 1999