SUPPLEMENTARY INFORMATION
Brain functional connectivity dynamics at rest in the aftermath of affective and cognitive challenges.
Julian Gaviria1, Gwladys Rey3, Thomas Bolton4,5, Jaime Delgado2 Dimitri Van de Ville4,5, Patrik Vuilleumier1,2,3
Supplementary methods Affective stimuli
Two videoclips with strong emotional content were used to modulate affective state, with a duration of five minutes each. One movie excerpt was edited from the film “21 Grams” [González Iñárritu, 2003], which has been used in previous mood induction studies (Hanich et al., 2014; Shiota &
Levenson, 2010) and more recently in Borchardt et al. (2017). The other clip was edited from Sophie's Choice [Pakula, 1981], which has also been validated by previous studies in terms of valence, arousal, and type of emotions [Raz et al., 2012; Raz et al., 2016]. These clips were selected based on several similarities regarding their content and characters, with the purpose of controlling - as best as possible - the psychological effects of naturalistic emotional stimuli. Both excerpts included young mothers as main characters, experiencing the loss of their underage children. Participants were instructed to feel as involved as possible in the drama of the movies. In addition, two neutral clips of 5 minutes each were taken from documentary interviews or TV pieces freely available in the internet, with verbally interacting people generally similar to negative movies but without any emotional aspect. The free software iMovie (https://www.apple.com/lae/imovie/) was used for editing the movies.
This resulted in a set of 4 video clips that were validated in a preliminary pilot study, in which a different group of healthy volunteers (n=28) rated the pleasantness, intensity, and type of emotions elicited by each movie. The final clips were compared in terms of low-level features (sound level, luminance, spatial frequency and motion), and no significant differences between the negative and neutral conditions were found.
Psychological measures
Affect questionnaires assessing the participant’s subjective emotional state were given before and after each experimental context (i.e., one rating for the neutral context, and another for the negative context), including the state scale of the Positive and Negative Affect Scales, PANAS (Watson et al., 1988). Participants also rated their affective perception of the content of each movie with a 9-point Likert scale combined with the self-assessment Manikin, SAM [Bradley and Lang, 1994], as used in previous work (Borchardt et al., 2017; Hanich et al., 2014). These ratings included hedonic experience (“How pleasant or unpleasant was your own experience of the scene?”,1 = very unpleasant, 9 = very pleasant). ( S1A), valence (“How happy/pleased or unhappy/dissatisfied are you?”, 1 = very unhappy, 9 = very happy), and arousal (“How awake/aroused or calm/drowsy do you feel?” ,1 = very calm, 9 = very aroused).
Cognitive control tasks
To probe for the effect of cognitive control demands on subsequent resting state and its modulation by the negative affect, we used two different cognitive tasks known to demand similar executive resources [Schuch and Koch, 2014]. Each participant performed these two tasks, one in the “neutral context”, and the other in the “negative context”, respectively. One task was a Flanker-number paradigm with digits as stimuli (Eriksen and Eriksen 1974; Hübner et al. 2010; Schuch and Koch 2014), and the second a Stroop-like paradigm with superimposed face–name stimuli [Egner and Hirsch, 2005; Schuch and Koch, 2014; Stroop, 1935]. Our choice of these tasks was guided by several factors and backed up by careful pilot testing. First, previous studies reported high sensibility of both tasks to negative emotional priming [Egner, 2007; Hommel et al., 2004; Mayr et al., 2003; Mayr and Awh, 2009; Schuch and Koch, 2014]. Second, both tasks produce reliable response conflicts, with the possibility to present many trials, with minimal learning, and both allow measuring conflict adaptation effects on RTs across successive trials [Macleod, 1991; Verguts et al., 2011]. Third, our
pilot results yielded very similar interference effects in both tasks, with and without emotional modulation (see “Results” section).
These two cognitive tasks (face Stroop task and number Flanker task) adjusted for fMRI compatibility.
Each task comprised 80 trials (40 congruent and 40 incongruent). Individual stimulus presentation lasted 1s, followed by an inter-trial interval with a central fixation cross randomly jittered from 2 to 4.9s. Both tasks were counterbalanced across participants. Each task combined congruent (C) and incongruent (I) trials in which a central target (name in Stroop, number in Flanker) was presented for a binary classification response (male/female for Stroop, odd/even for Flanker), together with a distractor corresponding to either the same (C) or opposite (I) response. In addition, a given trial could be preceded by either the same or opposite condition (C or I trials), in a semi-random but balanced order. This resulted in 4 trial types, allowing subsequent behavioral analysis according to both current compatibility (indicated by upper-case letters C and I) and compatibility of the preceding trial (indicated by lower-case letters c and i) (see Fig. 1B for a schematic description).
Special attention was given to balance the difficulty and performance of both tasks during extensive pilot testing.
Face-name Stroop task
All face images in the Stroop task were obtained from the Karolinska Directed Emotional Faces, KDEF [Calvo and Lundqvist, 2008]. Participants saw pictures of faces superimposed by names. The name and face corresponded to the same gender in 50 % of the trials (congruent stimuli). We used images of 35 male and 35 female faces (portrait format, size 6. cm x 8.5 cm). To obtain optimal homogeneity of visual features, all images were standardized in terms of color background (grey), color portray (black & white), and face proportion image (close-up). We used 70 different names from the French language whose unambiguous gender association was verified tested during preliminary piloting. All stimuli appeared in the center of a screen outside the scanner bore, reflected on a mirror placed on the head coil. Participants gave responses with an MRI-compatible device with two key buttons at
the right side. The instruction was to categorize the name as male or female. The stimuli size and screen-participant distance were reduced in a 1:2 ratio (approx.) for the trials outside the scanner.
Number Flanker task
The stimuli were digits 1, 2, 3, 4, 6, 7, 8, and 9, presented on gray background. They were displayed with the same font size and color as names in the Stroop task. To control for negative priming, a digit could be repeated as a flanker not less than nine trials after being presented as target. The target digit appeared in the center of the screen, reflected on the head-coil mirror. Two flanker digits were presented to the left and right of the target, at a distance of 20 mm from the target. The left and right flanker digits were always the same. The subjects’ task was to categorize the central digit as odd or even, and to respond by pressing a left or right response key from the MRI-compatible device. The stimuli size and screen-participant distance were reduced in a 1:2 ratio (approx.) for the trials’ tasks given inside vs outside the scanner.
FMRI data acquisition
Neuroimaging data were collected using a 3T Magnetom TIM Trio scanner (Siemens, Germany) and a 32 channels head-coil. The Blood Oxygenation Level Dependent (BOLD) contrast was measured using a T2*-weighted echo-planar sequence (EPI). Five hundred functional volumes of 36 axial slices each (TR/TE/flip angle = 1300ms/30ms/80°, FOV=192 mm, resolution=64×64, isotropic voxels of 3.2 mm3, distance factor 20%) were acquired in one single continuous scanning run. We collected a high- resolution T1-weighted anatomical image (TR/TI/TE/flip angle=1900ms/900ms/2.27ms/9°, FOV=230mm, resolution=256×256, slice thickness=0.9mm, 192 sagittal slices) at the end of the first session.
Preprocessing for GLM-based fMRI analysis on videoclips and tasks
Standard image preprocessing procedures were applied using SPM12 (www.fil.ion.ucl.ac.uk/spm/).
Functional images from the video clips and the tasks were realigned, slice-time corrected, normalized
and co-registered to individual skull stripped anatomical images comprised by the probabilistic maps of CSF (cerebrospinal fluid), gray and white matter extracted by the DARTEL Algorithm [Ashburner, 2007]. Data was spatially smoothed with an 8 mm Gaussian kernel.
GLM-based fMRI analysis
Preprocessed fMRI data from the video clips, cognitive tasks, and resting blocks were separately analyzed using the General Linear Model as implemented in SPM12. For the movie analysis, four covariates of interest were used in the regression analysis to model each clip. Regressors in these models were convolved with a standard hemodynamic response function (HRF) according to a blocked design, which was then submitted to a univariate regression analysis. Realignment parameters were also added to the design matrices of both models, to account for any residual movement confounds. In all cases, the design matrix included low-frequency drifts (cutoff frequency at 1/128 Hz). Flexible factorial analyses of variance (ANOVAs) were performed on the main contrasts of interest. All statistical analyses were carried out at the whole-brain level, with a threshold at P <
0.05, FWE corrected at whole brain level (unless specified otherwise).
Preprocessing for co-activation pattern (CAP) analysis on resting periods
Following Bolton et al. [Bolton et al., 2020], standard image preprocessing procedures were applied using the DPABI toolbox [Yan et al., 2016] . Functional images from the resting periods were realigned, slice-time corrected, normalized and co-registered to individual skull-stripped anatomical images defined by probabilistic maps of CSF (cerebrospinal fluid), gray and white matter. Data was spatially smoothed with a 5 mm Gaussian kernel. Time-series from ROIs in the white matter and cerebrospinal fluid, plus the six affine motion parameters from realignment (including their respective first-order derivatives), were used as nuisance variables to be regressed out from the data.
No global signal regression was carried out given the absence of agreement in the field, particularly on this regard [Murphy and Fox, 2017]. Furthermore, the data were band-pass filtered between 0.01 and 0.10 Hz. All image volumes with frame-wise displacement above 0.5 mm were discarded, as well
as subjects with more than 50% of scrubbed frames [Power et al., 2014]. Less than 5% of the total time frames were scrubbed, and estimated with cubic spline interpolation, in order to avoid alterations in the signal’s temporal continuity, and producing additional artifacts [Chen et al., 2017;
Hutchison et al., 2013]. However, removing corrupted frames for executing further analyses results less sensitive in the CAPs methodology compared to others dFC methods, due to the CAPs’ single- volume temporal resolution and minimal set of assumptions [Liu et al., 2018].
Supplementary Results
Psychological indices of affective stimulation
Subjective affective ratings of the content of negative movies (Likert scale from 1 to 9) yielded significant differences relative to neutral movies. Valence was more negative for negative
than neutral clips [F (1, 72) = 3.97, p < .01] ( Fig. S1A), but there was no difference between movies within the same emotional conditions (neutral, p = .66; negative affect, p = .84). Arousal was also higher for the negative than neutral movies [F (1, 72) = 3.96, p < .01]; again, no differences within each emotional condition (neutral, p = 1; negative, p = 0.8). Furthermore, there was a significant difference in subjective affect before and after exposure to movies in the “negative” context, with more negative scores on PANAS scale in the post scan ratings[t (18) = 1.7, p < 0.01], and consequently, less positive scores as well [t (18) > 1.7, p < 0.05] (see Figure S1. B.). Such differences were not observed for the “neutral” context [t (18) > 1.7, p > 0.1] ( Fig. S1B).
GLM-based fMRI analysis of emotion effects during movie clips
Brain activity during movie watching was assessed for completeness to verify the effectiveness of our protocol. Differential responses to negative vs. neutral movies was examined by direct contrasts between these conditions. As expected, higher activation (FWE p<.05) was observed in bilateral visual areas, orbital and lateral prefrontal areas, insula, and precuneus during the negative movies (Table S1). To note, our experimental design did not aim at capturing detailed changes in neural activity during movies themselves, given our focus on resting periods and the complex fluctuating
nature of the fMRI-BOLD signal observed during naturalistic stimulation over long time-windows such as with our movie clips [Hutchison et al., 2013; Simony et al., 2016]. Conversely, the neutral movies evoked greater activation in middle occipital, angular, and bilateral superior frontal gyri (Table S1). Because the precuneus is a core region of the DMN activated at rest and, in addition, was modulated by the emotional content of movies (bilateral cluster with peak at 4 -52 54; p < .05, FWE- corrected for the whole brain level; k = 486), this region was subsequently used as the seed of our CAP analysis. Importantly, resting state blocks used for our CAP analysis are entirely different from movie-related data used to define these coordinates.
Behavioral indices of cognitive control across tasks
Detailed analyses of behavioral performance demonstrated consistent and similar effects in both the Flanker and Stroop tasks, confirming the successful validation and standardization of these paradigms obtained in our piloting testing. First, for interference costs due to attentional conflict (congruency effect, i.e., I vs C trials), there was no significant differences between the RT of congruent (t (18)= 1.73, P < 1) or incongruent trials (t (18)= 1.73, P > 1), when both tasks were compared. Second, Pearson correlation analyses denoted a significant and similar statistical relationship between both tasks across participants for the response times (RTs) in both the congruent (r=.9, p<.0001) and incongruent trials (r=.8, p=.0001). Third, an omnibus LMM-based ANOVA with factors of “task” (levels: Stroop and Flanker), and “trial” (levels: cC, iC, iI, cI) (see Fig. 1B) revealed no main effect of task [F(1,126)= 3.36, P=.08)], and no significant interaction between the task and type of trial ([F (3,126) = 0.9, p = .45]. Also, post-hoc analyses revealed no significant variance for any of the different trial types when both tasks were directly compared [cC trials: β=
9.46; t(126)=1.0, p= 0.3; iC trials: β= 4.23, t(126)=0.4; p= 0.7; iI trials: β= 14.9, t(126)= 1.6, p= 0.15; cI trials: β= 14.3, t(126)= 1.5, p= 0.16]. Finally, the magnitude of interference costs on RTs (I minus C) did not differ between the two tasks [F (1,18) = 1.3, p = .5]; see Fig. S2A., and S2B.).
Impact of negative affect on cognitive control.
As described in the main text, interference costs were reliably observed in RTs (I > C) across the two cognitive control tasks and significantly modulated by emotional context. We also computed the magnitude of the congruency sequence effect (CSE; Egner 2007), reflecting differences in RTs on the four type of trials (see Fig. 1B) in these cognitive tasks for both experimental contexts (i.e., negative affect and neutral). The RT data were submitted to a LMM-ANOVA in which affective contexts were represented by a two-levels factor (negative vs neutral) and trial type was modeled by two separate two-levels factors [current trial type (xC and xI), and previous trial type (Cx, Ix)]. This analysis yielded a significant interaction between current and previous trial types (F (1,126) = 46.1, p < 0.001), and between current trial type and emotion context (F (1,126) = 4.9, p < 0.03); but no triple interaction (F (1,18) = .002, p <.9). Nevertheless, a post-hoc joint test revealed significant differences only for the most difficult type of trials (cI), when cognitive task performance was directly compared between the two affective contexts for each trial type (negative vs neutral, see Table S2). Thus, our behavioral data from cognitive control tasks fully accord with previous research [Gendolla et al., 2015; Schuch and Koch, 2014] and further validate the effectiveness of our protocol.
Neural effects of cognitive control tasks.
Finally, for completeness, we also verified that our cognitive control tasks activated brain networks previously reported in similar paradigms [Botvinick et al., 2001; Egner, 2007; Egner et al., 2008; Egner and Hirsch, 2005; Kerns, 2004; Rey et al., 2014]. To do so, we compared fMRI activity during the main attentional conditions of both tasks (Stroop and Flanker tasks combined) using a standard GLM analysis for event-related designs with SPM [Josephs et al., 1997]. As expected, we found robust congruency effects associated with interference costs (I>C trials) in several brain areas involved in conflict monitoring and selective attention. These activations encompassed posterior parietal areas, dorsolateral prefrontal cortex, and extrastriate visual areas bilaterally (all p<.05. FWE corrected at cluster level), overlapping with the dorsal attention network (DAN). Altogether, both behavioral and
neural data clearly demonstrate a reliable engagement of attentional control systems during our cognitive tasks.
Captions for Supplementary Figures
Figure S1. Behavioral measures. A) Subjective affective ratings of the movies obtained after fMRI scanning, assessing emotional experience, valence, and arousal felt for each movie clip. B) Mean scores of subjective negative and positive affect measures before (pre-experimental score) and after (post- experimental context) scanning, for both affective contexts (negative and neutral) respectively. Whiskers stand for standard deviation of means. * p>0.05;
** p>0.01; ***p>0.001.
Figure S2. Behavioral results from cognitive control tasks. A. A robust interference cost (longer reaction times on incongruent vs congruent trials) was observed, with highly similar performance in the two tasks (Stroop and Flanker).
B. Reaction Times were the longest on the most difficult task condition with incongruent trials preceded by congruent trials (cI), compared to congruent trials preceded by another congruent trial (cC), again with no significant difference between the two tasks. C. The interference effect was present after exposure to both negative and neutral movies, but enhanced in the former “negative”
context, as compared with the latter “neutral” context. D. Negative emotion induced selective increases in RTs for the most difficult trial type (cI trials), relative to the neutral context. No significant difference was produced by the negative emotional context on RTs for the easiest (cC congruent) trials. **
p>0.01; ***p>0.001.
Figure S3 Preliminary CAP analysis . Illustration of the eight brain networks identified by our CAP cluster analysis and co-activating with the precuneus during rest periods. B) Spatial maps for each CAP corresponding to distinct functional networks coupled with the precuneus across time. C) Occurrence rates for each CAP denoting their temporal fluctuation across the different resting conditions. Statistical analysis of these occurrence rates (one-way) revealed that 3 out of the 8 clusters presented significant differences between conditions. P values were adjusted by the FDR method for multiple testing under dependency.
* p>0.05
Figure S4. Functional relationship of CAP expression with behavioural indices. Up. Probability (P) values of partial associations between affective scores (PANAS) and the three CAPs of interest. PANAS ratings measured before (pre) and after (post) the neutral (A) and negative (B) contexts, respectively, were modeled as predictors. Entry rates of these CAPs during the “neutral movie rest1” (A) and “negative movie rest1” (B) conditions were introduced as outcome variables. Bottom. Probability (P) values of partial associations between cognitive control performance and relevant CAPs, using RT measures across trials as predictors and CAP expression during the “movie+task rest2” conditions as outcomes. Each matrix illustrates the significance values for all possible associations in the neutral (C) and negative contexts (D), respectively, to underscore the high selectivity of significant relationships between the relevant CAPs and specific behavioural indices. For completeness, we also verified that none of the other CAPs showed any even near significant association with these behavioural measures (data not illustrated). P values were adjusted by the FDR method for multiple testing under dependency.
A.
C.
B.
D.
Supplementary Tables
Table S1. Brain activation to negative movie clips vs. neutral clips.
Clusters are listed using a threshold at p <0.001 uncorrected with *p < 0.05, FWE-corrected peaks.
Coordinates (MNI)
Brain Region x y Z K Z-score
Neutral > Negative affect
Angular gyrus -44 -64 32 162 4.88
Middle Occipital G. 42 -74 36 322 370
Superior Frontal G. 22 32 42 122 4.05
Superior Frontal
Gyrus -24 24 42 177 4.22
Negative affect > Neutral Inferior occipital
gyrus* 52 -70 2 839 9.66
Inferior occipital
gyrus* -46 -78 6 784 7.75
Precuneus* 4 -52 54 486 6.50
Lingual gyrus* 10 -78 -6 309 6.20
Lateral orbitofrontal
cortex -36 26 -6 609 4.17
Middle frontal gyrus -40 52 18 176 3.82
Anterior Insula -40 14 -4 84 3.79
Putamen 20 4 -6 66 3.24
Table S2. Mixed model analysis of reaction times in the cognitive control tasks. Eight regressors were included in this model, representing the type of trial (cC=previous congruent, current congruent; iC=previous incongruent, current congruent; iI= previous incongruent, current incongruent; cI= previous congruent, current incongruent), and the emotional context (neutral, negative) as fixed effects. Scores from each participant were included as random effects.
Standardized coefficient estimates from each condition (N_xC=congruent trials in the neutral context; N_xI=incongruent trials in the neutral context; A_xC=
congruent trials in the negative context; A_xI= incongruent trials in the negative context) were contrasted to assess the influence of negative affect on cognitive control performance. P value adjustment: ). P-adjustment by the FDR method for multiple testing under dependency.
ANOVA Estimates β (SE) LS means
Curent trial Previous x Curent
Emotion
x Curent Context cC iC iI cI Contrast CIlow CIhigh P
value
F(1,126)=123 .7 p<.001
F(1,126)=19.
9 p< .001
F(1,126)=4.9 p< .05
Neutr al (N)
567.
6 (14.5
)
598.
8 (14.5
)
619.
2 (14.5
)
629.
9 (14.5
)
A cC > N
cC -36,2 28,1 0,99
A iC > N
iC -20,9 43,4 0,92
Negative affect (A) 571.7
(14.5) 587.5
(14.5) 619.2 (14.5) 654.1
(14.5)
A cI > N
cI -56,3 8,01 0,03
A iI > N
iI -41,8 22,5 0,19
Table S3. Brain activation during cognitive control tasks. Top part. Brain regions with increased activity in the contrast comparing congruent >
incongruent trials. Bottom part. Brain regions with increased activity in the contrast comparing incongruent > congruent trials (p < .001, FWE-corrected at cluster level).
Coordinates (MNI)
Brain Region x y z K Z-score
Congruent > Incongruent
Hippocampus -22 -38 -4 119 5.91
Incongruent > Congruent
Superior Parietal lobe -22 -62 46 784 8.04
Superior Parietal lobe 28 -68 30 675 5.39
Middle frontal Gyrus -24 4 48 188 6.33
Supplementary Motor
C. -80 20 48 250 5.43
Inferior temporal Gyrus -48 -66 -16 372 5.07
Inferior temporal Gyrus 46 -62 -14 129 5.01
Middle Occipital Gyrus 36 -82 12 287 5.29
Middle frontal gyrus -40 26 22 140 5.81
Middle frontal gyrus 42 10 34 126 5.08
Putamen -22 0 4 118 4.73
Putamen 28 8 4 328 5.49
Table S4. Anatomical peaks in spatial maps of CAP3, CAP4, and CAP7.
Regions with positive z-score co-activated with the time-course of the precuneus seed at Z>1.7 (p<.05), whereas regions with negative z-score co-deactivated at Z =< -1.7 (p<.05).
Region (MNI) x Y z K> Z max.
CAP 3
Angular gyrus -54 -68 16 1226 4.1
Angular gyrus 56 -62 28 899 3.5
Medial frontal gyrus -2 54 -16 715 3.5
Insula -40 2 2 297 -2.2
Insula 36 18 4 736 -2.5
Supramarginal gyrus -64 -26 30 552 -2.8
Supramarginal gyrus 62 -28 40 736 -2.8
Middle cingulate gyrus 8 12 38 58 -1.7
CAP 4
Superior frontal gyrus -22 -2 64 699 2.8
Superior frontal gyrus 28 6 62 716 3.2
Precuneus 12 -70 54 435 3.3
Precuneus -8 -68 56 693 2.9
Supramarginal gyrus -58 -32 40 725 3.3
Supramarginal gyrus 60 -26 38 593 3.4
Middle cingulate gyrus 4 10 42 85 2.3
Middle frontal gyrus -36 52 18 541 2.5
Middle frontal gyrus 36 44 30 357 2.3
Posterior cingulate gyrus -4 -52 28 153 -2.1
Angular gyrus -48 -64 38 587 -2.4
Angular gyrus 54 -62 38 60 -2.1
Orbital med. frontal gyrus -2 28 -30 114 -1.9
Orbital inferior frontal gyrus -44 34 -14 21 -2.1
Middle temporal gyrus -60 -42 -2 235 -2.1
Inferior temporal gyrus -62 -10 -22 208 -2.1
CAP 7
Postcentral gyrus -30 -34 68 2627 3.7
Postcentral gyrus 32 -36 62 2341 3.4
Précentral gyrus -32 -24 64 3.7
Precentral gyrus 34 -24 64 3.4
Supplementary motor area -4 18 50 3.7
Supplementary motor area 10 4 42 3.4
Middle cingulate gyrus 2 -10 48 50 2.1
Thalamus 4 -22 2 58 -2.5
Table S5. Results of the 2x2 factorial ANOVA with LMMs on temporal parameters of relevant CAPs. The expression of each CAP was computed in terms of occurrences, entries, and durations. They were separately tested for the main effects of emotional context (negative vs neutral) and preceding event type (movie alone vs movie + cognitive task), across the 4 corresponding resting conditions (two-way analysis, 2x2 levels). Abbreviations: B= Baseline rest; N=
neutral movie rest1; NT= neutral movie+task rest2; A= negative movie rest1;
AT= neg rest2. Reported differences are indicated by. p>0.1; * p>0.05; **
p>0.01; ***p>0.001). P-adjustment by the FDR method for multiple testing under dependency
ANOVA Estimates β (SE)
CAP Context Event Context
x Event N NT A AT
CAP3
OCCURRENCES F(1,54)=5.8.
p = .01 (1,54)=2.4
p = .12 (1,54)=1.3.
p = .26 12.5 (2.0) 11.8
(2.0) 18.5 (2.0) 14.0
(2.0)
CAP3 ENTRIES
F(1,54)=8.0.
p = .01
F(1,54)=0.8 p = .36
F(1,54)=2.9.
p = .09
11.3 (1.3)
13.8 (1.3)
17.3 (1.3)
13.8 (1.3)
CAP4
DURATION F(1,54)=7.6.
p = .01 F(1,54)=0.04
p = .98 F(1,54)=2.5.
p = .11 3.1
(0.3) 2.6 (.3) 3.5
(0.3) 4.1 (0.3)
CAP7
OCCURRENCES F(1,54)=10.6.
p = .00 (1,54)=1.3
p = .24 (1,54)=1.3.
p = .90 11.2
(2.6) 13.0 (2.6) 5.5
(2.6) 7.7 (2.6)
CAP7
ENTRIES F(1,54)=13.5.
p =.00 F(1,54)=2.4
p = .12 F(1,54)=1.1.
p = .30 12.1 (2.2) 12.9
(2.2) 5.6 (2.2) 9.2
(2.2)
