Accepted Manuscript Adult ADHD and Working Memory: Neural Evidence of Impaired Encoding Soyeon Kim, Zhongxu Liu, Daniel Glizer, Rosemary Tannock, Steven Woltering PII: DOI: Reference:
S1388-2457(13)01348-5 http://dx.doi.org/10.1016/j.clinph.2013.12.094 CLINPH 2006941
To appear in:
Clinical Neurophysiology
Accepted Date:
4 December 2013
Please cite this article as: Kim, S., Liu, Z., Glizer, D., Tannock, R., Woltering, S., Adult ADHD and Working Memory: Neural Evidence of Impaired Encoding, Clinical Neurophysiology (2013), doi: http://dx.doi.org/10.1016/ j.clinph.2013.12.094
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
WM encoding in ADHD
Adult ADHD and Working Memory: Neural Evidence of Impaired Encoding
1 2 3 4
Soyeon Kim, Zhongxu Liu, Daniel Glizer, Rosemary Tannock*, Steven Woltering*
5 6
Department of Applied Psychology & Human Development, OISE, University of Toronto,
7
252 Bloor Street West, Toronto, Ont., Canada, M5S 1V6
8 9
E-mails: Soyeon Kim (
[email protected]), Zhong-Xu Liu (
[email protected]),
10
Daniel Glizer (
[email protected]), Rosemary Tannock (
[email protected]),
11
StevenWoltering (
[email protected]).
12 13
* Corresponding authors:
14
Steven Woltering
15
Department of Applied Psychology & Human Development, OISE, University of Toronto, 252
16
Bloor Street West, Toronto, Ont. Canada, M5S 1V6
17
Tel: +1-416-978-0923
18
[email protected] 19 20
or
21 22
Rosemary Tannock
23
Department of Applied Psychology & Human Development, OISE, University of Toronto, 252
24
Bloor Street West, Toronto, Ont. Canada, M5S 1V6
25
E-mail:
[email protected] 26 27 28 1
WM encoding in ADHD
29 30
Highlights
31 32
ADHD college students are compared to their typically developing peers on neural and behavioural indices of working memory (WM).
33 34
Neural indices of working memory (P3) are examined during working memory encoding using EEG
35 36
ADHD group showed lower P3 during working memory encoding, suggesting an inefficient encoding process in WM.
37 38 39 40 41 42 43 44
2
WM encoding in ADHD
45 46
Abstract
47
Objectives: To investigate neural and behavioural correlates of visual encoding during a working
48
memory (WM) task in young adults with and without Attention-Deficit/Hyperactivity Disorder (ADHD).
49
Methods: A sample of 30 college students currently meeting a diagnosis of ADHD and 25 typically
50
developing students, matched on age and gender, performed a delayed match-to-sample task with low
51
and high memory load conditions. Dense-array electroencephalography was recorded. Specifically, the
52
P3, an event related potential (ERP) associated with WM, was examined because of its relation with
53
attentional allocation during WM. Task performance (accuracy, reaction time) as well as performance
54
on other neuropsychological tasks of WM was analyzed.
55
Results: Neural differences were found between the groups. Specifically, the P3 amplitude was smaller
56
in the ADHD group compared to the comparison group for both load conditions at parietal-occipital
57
sites. Lower scores on behavioral working memory tasks were suggestive of impaired behavioural WM
58
performance in the ADHD group.
59
Conclusions: Findings from this study provide the first evidence of neural differences in the encoding
60
stage of WM in young adults with ADHD, suggesting ineffective allocation of attentional resources
61
involved in encoding of information in WM.
62
Significance: These findings, reflecting alternate neural functioning of WM, may explain some of the
63
difficulties related to WM functioning that college students with ADHD report in their every day
64
cognitive functioning.
65 66 67
Keywords: Working memory, P3, adults with ADHD, encoding.
68 69
3
WM encoding in ADHD
70
INTRODUCTION
71
Attention-deficit/Hyperactivity Disorder (ADHD) is the most frequently diagnosed psychiatric
72
disorder in childhood with worldwide prevalence rates estimated at 5.3% (Polanczyk et al., 2007).
73
Longitudinal studies show that approximately 65% of children with ADHD continue to show symptoms
74
in adulthood (Barkley et al. 2002; Faraone et al., 2006; Polanczyk and Rhode, 2007). A newly-emergent
75
subset of young adults with ADHD – those who gain entrance into post-secondary education – is of
76
interest because of their ongoing impairments despite educational success relative to others with ADHD
77
(DuPaul et al., 2009). These students present various impairments in social and academic domains. For
78
instance, they exhibit difficulties with college adjustment, social skills and self-esteem (Shaw-Zirt et al.,
79
2005), lower Grade Point Averages, poorer academic success, and they are more likely to be placed on
80
academic probation (Heiligenstein, et al., 1999; Frazier, et al., 2007). Working memory (WM), seen as
81
one of the core deficits of ADHD (Martinussen, et al., 2005; Hervey et al., 2004), could be a key factor
82
explaining some of the functional impairments mentioned above. WM allows us to temporarily hold and
83
manipulate information “on-line” for a few seconds in order to respond effectively based on that
84
information (Baddeley, 2003). WM involves several stages of information processing: encoding,
85
maintenance (rehearsal) and retrieval. WM has been linked with many real world activities (e.g., reading
86
comprehension, following a conversation, problem-solving; Baddeley and Hitch, 1994) and therefore
87
WM impairments may have pernicious effects on everyday life, especially considering the social and
88
academic demands students face in college.
89
WM processing can be effectively investigated using ERP methodology due to its fine temporal
90
resolution. Among ERP components, changes in P3 amplitudes, occurring about 300 ms after the
91
stimulus, have been reliably associated with WM functioning. Several theories have been proposed to
92
interpret the role of the P3 in working memory tasks (for overview, see Kok, 2001; Polich, 2007),
93
however, there is no consensus as to what the P3 represents in the context of a WM task, especially
94
during the encoding phase. For example, Donchin and Coles (1988) interpreted the P3 within the
4
WM encoding in ADHD
95
framework of a "context updating theory" which proposes that P3 is increased when there is a need to
96
update, change, or refresh the current contents of WM. Researchers also focus on the intricate role of
97
attentional processes in working memory functioning. For instance, some suggest that WM capacity is
98
determined by an executive attentional process (Kane et al., 2007; Vogel, 2005), or the „capacity of the
99
focus of attention‟ (i.e. the scope of attention; Cowan, 2001). Ford, White, Lim, and Pfefferbaum (1994)
100
suggested that P3 amplitude reflects the effort to allocate attention (see also, Polich, 2007). Similarly,
101
Kok (2001) proposed that the P3 reflects the „attentional capacity invested in the categorization of task
102
relevant events‟. In sum, these studies suggest a possible link between P3 amplitude and allocation of
103
attentional resources that subserve working memory functioning. Thus, in the current study, we interpret
104
P3 activation as reflecting the direction of attentional resources towards encoding information into WM,
105
such as categorizing events or updating mental representations. However, attention acts at many levels
106
of processing during working memory, including early perceptual processes (Rutman et al., 2010). For
107
instance, ERP components such as the P1 are believed to represent sensory responses elicited by visual
108
stimuli as early as 100 ms (Luck, 2005). Thus, we also measure the P1 component to ascertain whether
109
perceptual/sensory processes are intact or altered in college students with ADHD.
110
A recent meta-analysis of six studies investigating the P3 component in adults with ADHD found
111
significantly reduced P3 amplitude compared to controls (Szuromi et al., 2011). However, all of the
112
studies used a GoNoGo paradigm: none used a WM task. Moreover, a recently published EEG study of
113
WM functioning in adult ADHD utilized an n-back task and mainly investigated time-frequency
114
measures (Missonier et al., 2013). Among other findings in that study, Missonier and colleagues (2013)
115
reported reduced low frequency oscillations (e.g., theta) in the ADHD group during the period in which
116
the P3 would have occurred (e.g., between 200-500 ms), suggesting that both time-and frequency
117
domain measures reflect similar underlying neural processes.
118
Only one study to date has investigated the P3 component in the context of a „delayed match-to-
119
sample‟ WM paradigm and this involved a comparison of participants with and without Schizophrenia
5
WM encoding in ADHD
120
(Haenschel et al., 2007). In this paradigm, the to-be-remembered target stimuli were presented in a
121
sequence that varied in length (e.g., 2 versus 3 stimuli) to measure effects of WM load. Longer
122
sequences place greater demands on WM in terms of storage and updating information compared to
123
short sequences. Then, after a delay period, the participant is asked whether a test-stimulus matched one
124
of the targets. The task uses irregular shapes as stimuli to avoid verbal/semantic processing.
125
Significantly reduced mean P1 (an early visual component) and P3 amplitudes were found in individuals
126
with Schizophrenia compared to controls. The load effect was present only for the P1 component. The
127
authors interpreted the reduced P1 activity as an early visual processing deficit whereas the decreased
128
P3 amplitudes were interpreted as evidence of reduced ability in categorizing or evaluating stimuli.
129
To investigate WM, the present study adapted Haenschel‟s WM paradigm (Haenschel et al.,
130
2007) and used it with a sample of college students with ADHD. This paradigm allowed us to
131
investigate the neural effects of encoding stimuli presented in sequence under different load conditions
132
(low versus high load). Specifically, the current task required participants to allocate attention to the
133
stimulus, register it, and then quickly shift attention to the next stimulus while keeping the previous one
134
in mind. This process requires both earlier attention allocation to sensory/perception of the stimuli and
135
later allocation of attention to evaluate the stimulus and update the mental representation. To the best of
136
our knowledge, no previous study has investigated the encoding phase of WM using a delayed-match-
137
to-sample task in participants with ADHD. A complete understanding of ADHD necessitates the study
138
of different subpopulations of this clinical condition, which represent different levels of functioning.
139
Our sample is unique in that it consists of college students with ADHD who are relatively high
140
functioning and successful academically, but who continue to show impairment (DuPaul et al., 2009). A
141
better understanding of mechanisms underlying their difficulties may inform interventions designed to
142
increase academic achievement and educational attainment in this population.
143 144
To determine whether participants with ADHD are impaired during WM encoding, we measured P1 and P3 amplitudes, which allowed us to ascertain at which stage (early vs. later visual processing)
6
WM encoding in ADHD
145
any impairment occurred. Furthermore, to test whether any impairment might vary according to the
146
level of WM demand, we utilized two load conditions (low, high) in the delayed match-to-sample WM
147
task. We predicted that individuals with ADHD would show impaired WM performance and reduced P3
148
amplitude and that these group differences would be most pronounced under the high-load WM
149
condition.
150 151
METHOD
152
1) Participants:
153
A total of 32 unmedicated individuals with ADHD (47% male; aged 19-35) and 25 control
154
participants (44% male; aged 19-32) participated in the study. Participants with ADHD were recruited
155
from University Student Disability Services in a major urban area, via email lists and flyers. Inclusion
156
criteria were; 1) current enrolment in a post-secondary program, 2) a previous diagnosis of ADHD, 3)
157
registration with respective university or college Student Disability Services, which requires
158
documented evidence of a previously confirmed diagnosis of ADHD (typically, but not invariably in
159
elementary school), and 4) aged 19-35. Exclusion criteria were; 1) uncorrected sensory impairment, 2)
160
major neurological dysfunction and psychosis, and 3) current use of sedating or mood altering
161
medication including stimulant medication for ADHD. Amongst the clinical sample, 4 reported
162
comorbid learning disability. Participants in the comparison group were recruited through campus
163
advertisements: they were required to have no history or current presentation of mental health disorders.
164
All participants completed the Adult ADHD Self-Report Scale (ASRS v1.1), the Symptom
165
Assessment-45 (SA-45; Maruish, 1999) and the Cognitive Failures Questionnaire (CFQ; Broadbent,
166
Cooper, FitzGerald & Parkes, 1982; Wallace et al., 2002) to quantify current symptoms of ADHD,
167
additional psychopathology, and associated levels of cognitive impairment, respectively. The ASRS is a
168
reliable and valid scale for evaluating current symptoms of ADHD in adults (Adler et al., 2006). The
169
ASRS v1.1 consists of eighteen questions based on the criteria used for diagnosing ADHD in the DSM-
7
WM encoding in ADHD
170
IV-TR. The SA–45 is a brief assessment of psychiatric symptom. It is based on the well-validated
171
longer version (SCL–90–R) to create a brief, yet thorough, measure of psychiatric symptoms. The CFQ
172
measures self-reported failures in perception, memory, and motor function in everyday life. This 25-
173
item questionnaire uses a 5-point Likert scale and has a good external validity (Broadbent, Cooper,
174
FitzGerald & Parkes, 1982; Wallace et al., 2002). Questions require participants to rank how often these
175
failures in cognition occur. Both ASRS and CFQ instruments yield a total score whereby a higher score
176
represents more ADHD symptoms and more cognitive failures, respectively (Table 1).
177 178
The study was approved by the Research Ethics Boards of the participating universities. All participants provided informed written consent before starting the study.
179 180
>
181 182
2) Tasks
183
Neuropsychological tasks measuring WM performance
184
Three neuropsychological tasks were used to measure WM performance. The Digit Span subtest from
185
the Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV) was used to assess verbal working
186
memory ability (Wechsler, 2008). The Spatial Span (SSP) and Spatial Working Memory (SWM) tasks
187
from the Cambridge Neuropsychological Testing Automated Battery (CANTAB) were used to assess
188
visual spatial WM. The SSP required participants to remember the spatial sequence of squares flashed
189
briefly one at a time on the screen. The scores represent the highest level at which the participant
190
reproduces at least one correct sequence. The SWM task requires participants to find a blue token in a
191
series of displayed boxes and use these to fill up an empty column, taking care not to return to boxes
192
where a blue token had been previously found. The standardized z-score for total number of errors for
193
the SWM was used.
194
8
WM encoding in ADHD
195 196
The ERP delayed match-to-sample task The delayed match-to-sample task was adapted from Haenschel et al (2007). Thirty-six different
197
abstract figures were used as stimuli. The task consisted of 2 WM load conditions (144 trials total): a
198
high load condition in which a sequence of 3 stimuli was presented (72 trials) and a low load condition
199
with a sequence of 2 stimuli (72 trials). The low load (2-stimulus) condition is shown in Figure 1). Our
200
task differed from that of Haenschel et al (2007) in that we made the task self-paced to better optimize
201
participants‟ readiness on computer-paced tasks which are known to be difficult in the ADHD
202
population (e.g., interference effects of „set shifting‟ between trials) and thereby obtain a purer measure
203
of working memory. Furthermore, we were also hoping to reduce the loss of trials due to artifacts.
204
Participants initiated each trial by pressing the space bar on a keyboard, which served to optimize their
205
readiness for the start of each trial. Trials were presented in blocks of 12 trials each of either a low or
206
high load. The blocks of high (3 Stimuli) and low (2 Stimuli) loads were randomly distributed. Before
207
each block, participants were shown which condition to expect (e.g., low or high load) and after each
208
block, feedback on accuracy was provided (as the percentage of trials that were correct). After the space
209
bar was pressed, a clear time of 600 ms was introduced before the fixation stimulus was presented for
210
400 ms. Each target stimulus appeared in the center of the screen for 600 ms (visual angle, 1.34) with an
211
inter-stimulus interval of 400 ms. Two seconds after the last target stimulus appeared (delay;
212
maintenance), a probe stimulus was shown for two seconds (target; retrieval: see figure 1). Participants
213
were asked to indicate whether the target stimulus was one of the previously presented stimuli by
214
pressing the keys labelled as „Same‟ or “Different” on the keypad using their dominant hand. Response
215
accuracy was emphasized, but to maintain a reasonable pace of processing, participants were required to
216
respond within a 2 second time frame, after which the trial would be terminated and scored as incorrect.
217
However, response time was recorded. E-prime 1.2 software (Psychology Software Tools, Inc.) was
218
used to control stimulus presentation and timing as well as to record the accuracy and latency of
219
responses.
9
WM encoding in ADHD
220
Trial time varied between 4-6 seconds for low load and 5-7 seconds for the high load. Since
221
between-trial inter-stimulus intervals would vary because of the self-pacing (participant pressed space bar
222
to start the next trial), total time for task completion was calculated and compared between groups. The
223
analysis indicated that there were no significant differences between the two groups in terms of total time
224
to complete the task (see table 2).
225 226
>
227 228 229
3) Neural data acquisition, processing: ERP data
230
The EEG was recorded with a 128-channel Geodesic Hydrocel Sensor Net at a 500 Hz sampling rate,
231
using Netstation stand-alone software (Electrical Geodesics Inc, Eugene, Oregon). Netstation was used
232
to filter (FIR, 0.5-30 Hz, 60 Hz notch) and segment the data (400 ms before stimulus onset and 2000
233
ms for low load, or 3000 ms for high load, after stimulus onset).
234
Segments containing artifacts were removed using automatic algorithms to detect eye blinks and
235
eye movements, as well as large drifts or spikes in the data. Eye blinks were detected when the vertical
236
eye channels exceeded a threshold of 150 µv (max-min) within a 160 ms (moving) time window
237
within each trial after running a 20 ms moving-average smoothing algorithm across the entire trial
238
period. Eye movements were detected when horizontal eye channels exceeded a threshold of 100 µv
239
(max-min) over a 200 ms time window. Channels were automatically marked bad when they exceeded
240
a transition threshold of 200 µv over the entire segment (max-min). Segments containing more than
241
20% bad channels were automatically removed. In addition, all epochs were visually inspected by
242
trained research assistants blind to the hypotheses. Bad channels were replaced by values interpolated
243
from neighbouring channel data using spherical splines. Results were verified manually by a research
244
assistant, who was unaware of the study objectives or hypotheses. Because we used a relatively low
10
WM encoding in ADHD
245
high-pass filter, a linear detrending algorithm was applied to individual epochs to eliminate any left-
246
over linear drift if it existed.
247
The data were then averaged for each subject and stimulus, average referenced, and coded
248
within the segment. The P1 components were scored as the maximum positivity between 80-160 ms
249
after stimulus-onset for each stimulus (e.g., 3 stimuli for high load) for electrode sites 75 (Oz), 70 (O1)
250
and 83 (O2). Similarly, P3 components were scored as the mean positivity between 250-500 ms after
251
stimulus onset for electrode sites 65, 66, 84, and 90 (See Supplementary Figure S1 for electrode map)
252
compared to the baseline. The mean activity was chosen to represent P3 because this is a more
253
elongated component with peaks (often several) that are less clearly defined compared to „perceptual‟
254
components (see also, Luck, 2005; Woltering et al., 2011). Electrodes were selected based on the
255
literature (e.g., see Polich, 2007, for rationale) as well as the maximal positivities in the grand average
256
waveforms (relative to the 200 ms baseline). Latency ranges were determined based on inspection of
257
the individual ERPs as well as Haenschel‟s et al., 2007 study. We acknowledge that, considering the
258
paucity of ERP studies using delayed-match-to-sample tasks, our component labeling may vary from
259
convention used in other tasks. Before exportation, data were baseline corrected for 200 ms prior to
260
stimulus onset, separately, for each individual stimulus. Only correct trials were used in the analysis.
261
Averages across electrode sites were used for the P1 and P3 analysis.
262 263
4) Data analysis:
264
Independent one-way ANOVAs were used to test for group differences in performance on the
265
neuropsychological tasks. Separate repeated measures ANOVAs were conducted to test for group
266
differences in accuracy and reaction time on the delayed match-to-sample task, with Load as the
267
repeated measure (High load vs. Low) and Group as a between participant factor (ADHD vs. Control).
268
By contrast, the incomplete factorial design of the Delayed-Match-To-Sample task (2 stimuli in Low
269
Load; 3 stimuli in High Load) precluded a Group (2) x Load (2) x (Stimulus; 2 or 3) repeated measures
11
WM encoding in ADHD
270
ANOVA to investigate effects on the P1 and P3 ERP components. Thus we used a multi-step approach
271
to disentangle the effects of Load, Stimulus, and Group on our ERP components (P1, P3). In Step-1, to
272
investigate the effects of Group and WM Load on the P3 and P1, we collapsed across Stimulus, and ran
273
a Group by Load Repeated Measures ANOVA. For Step-2 we ran a Group by Load Repeated Measures
274
ANOVA for the first and last Stimuli separately. The last stimuli of each load would represent the
275
situation in which a participant would hold the maximum amount of information in mind for each
276
condition when encoding new information. This analysis involved a comparison of the second (last)
277
stimulus of the low load with the third (last) stimulus of the high load: the second stimulus of the high
278
load was not compared to either the first or the second stimulus of the low load as it was not
279
conceptually equivalent to those stimuli. In Step-3, we ran separate Group by Stimulus repeated
280
measures ANOVAs for each load. For the high load, contrasts were run to test for differences between
281
the three stimuli. This step allowed us to investigate sequence effects of stimulus presentation, which
282
may provide insights in the effects of neural resources being allocated during encoding with increasing
283
amount of stimuli to be held in mind.
284
Two participants, whose ERP data yielded less than 10 trials, were excluded from this analysis.
285
Data points (behavioural and ERP) with SD's >3 were regarded as outliers and adjusted using the
286
winsorizing technique (Tabachnick and Fidell, 2001). Three data points in ADHD group and 2 data
287
points in control group were winsorized. No significant differences were found in trial-counts [Low
288
load: F (1, 49) = .258, p=.553, High load: F (1, 49) = .006, p=.938] between the groups.
289
Partial eta-squared values (η2) were computed to ascertain effect size (ES). According to Vacha-
290
Haase and Thompson (2004), ES based on η2 = .01 corresponds to a small effect, η2 = .10 corresponds to a
291
medium effect, and η2 = .25 represents a large effect.
292 293 294
RESULTS
12
WM encoding in ADHD
295 296
1. Behavioral Data Analysis of demographic and clinical variables confirmed that the ADHD and control groups
297
were comparable in terms of age [F (53) = 1.358, p=.25] and sex [χ² (1, 21) = .039, p = 1.00]. As
298
expected, ADHD participants reported significantly more ADHD symptoms [F (54) =1, p< .001] and
299
cognitive failures in everyday life than controls [F (42) = 59.91, p< .001]. The ADHD group also
300
reported more obsessive-compulsive symptoms [F (43) = 25.22, p< .001] on the SA-45, compared to the
301
Control group. Means and standard deviations are presented in Table 1.
302
Analysis of neuropsychological tests of WM performance revealed, as expected, that the ADHD
303
group obtained significantly lower scores on the CANTAB-SWM [F (1, 44) = 4.54, p = .039], compared
304
to the control group. ADHD group also showed poorer scores in WAIS Digit Span, albeit a non-
305
significant difference [F (1, 48) = 3.29, p = .076]. Groups did not differ on CANTAB-SSP. The mean
306
scores for the ADHD group on these two tests of WM fell in the low average range: their mean SWM z-
307
score was -.10, which corresponds to a scaled score of just below 7 (16th percentile); and their mean
308
Digit Span scaled score was 8.31 (25th percentile). By contrast, scores for the comparison group fell
309
well within the average range: mean Digit Span scaled score of 9.95 corresponds to the 50th percentile
310
and the SWM z-score was .51, which corresponds to a scaled score of 10 or the 50th percentile.
311
As predicted, analysis of the Delayed-Match-To-Sample Task, revealed a main effect of Load
312
on performance accuracy [F (1, 53) = 49.88, p = .000, ES = .49]. Specifically, all participants were less
313
accurate during the high WM load compared to the low WM load condition, indicating the success of
314
the experimental manipulation. The groups did not differ significantly in terms of either their accuracy
315
or response times (to the target stimulus). There was no significant Group-by-Load interaction for
316
either response accuracy or response times. Moreover, there were no group differences in terms of intra-
317
individual variability of response times and neither the main effect of Load nor the Group-by-Load
318
interaction was significant. Means and standard deviations of the neuropsychological tasks and the
319
delayed match to sample task measures are shown in Table 2.
13
WM encoding in ADHD
320 321
>
322 323 324
2) ERP data The waveforms for each stimulus are presented in Figure 2, in Panel A for the low load and Panel B
325
for the high load. Table 3 shows the means and standard deviations for the ERP components for each
326
load and for each group collapsed across stimulus. The waveforms and (difference) topoplots for the P3,
327
collapsed across stimulus, are presented in Figure 3, Panel A for the low load and Panel B for the high
328
load. As can be seen in the topoplots, the difference in P3 amplitude between the ADHD and control
329
group was observed primarily in the parietal/occipital regions (see Supplementary Figure S2 for a t-plot).
330 331 332
P1 Analyses For P1 amplitude, Step-1 analysis, which examined the effects of Group and Load collapsed
333
across stimuli, showed a significant main effect of Load [F (1, 53) = 8.26, p = .006, ES = .14] with
334
higher amplitude for high load. There was no main effect of Group or Group-by-Load interaction. Step-
335
2 analyzed Group and Load effects separately for the first and last stimuli. For the first stimulus there
336
was a significant Group-by-Load interaction [F (1, 53 = 6.27, p = .015, ES = .106] in the absence of
337
main effects for either Group or Load. Post-hoc tests revealed that participants with ADHD had
338
increased P1 amplitudes for the High compared to the Low load, whereas the comparison group did not
339
show this effect (p < .05). For the last stimulus, a main effect of Load was present [F (1, 52) = 5.76, p
340
= .02, ES = .10] but neither the main effect of Group nor the Group-by-Load interaction was significant.
341
Step-3 analysis of P1 during the low load condition revealed a main effect of Stimulus, [F (1, 53) = 9.31,
342
p = .004, ES = .15] with lower P1 amplitudes for the first stimulus compared to the last. No Group, or
343
Group-by-Stimulus interaction effect was found. The high load also revealed a main effect of Stimulus,
344
[F (1, 51) = 13.61, p = .000, ES = .35], with lower P1 amplitudes for the first Stimulus compared to the
14
WM encoding in ADHD
345
second and third. No Group or Group-by-Stimulus interaction was found. Last, the P1 did not correlate
346
with any clinical questionnaire or behavioral measures.
347
For P1 latency, Step-1 analysis revealed no main effect of Load, Group, or Load-by-Group
348
interaction for either the high or low load conditions. On the Step-2 analysis, no main effects of Group,
349
Load, or a Stimulus-by-Group interaction were found, nor were there any significant effects for the first
350
stimulus. On the Step-3 analysis, neither the main effects of Stimulus or Group, nor the Stimulus-by-
351
Group interaction was significant.
352 353 354
P3 Analyses For P3 amplitude, our Step-1 analysis showed a significant main effect of Group, [F (1, 53) =
355
6.30, p = .015, ES = .11], with the ADHD group showing smaller P3 amplitudes than the control group.
356
There was also a main effect of Load, [F (1, 53) = 4.04, p = .049, ES = .07], with larger P3 amplitudes
357
for the high load confirming that our P3 index was sensitive to effects of load. No Group-by-Load
358
interaction was found.
359
The Step-2 analysis tested the Group by Load effects separately for the first and last Stimuli.
360
There was a main effect of Group for the first, [F (1, 53) = 5.31, p = .025, ES = .09], and last [F (1, 52)
361
= 5.76, p = .020, ES = .10], stimuli suggesting the group difference was present across stimuli. There
362
was a main effect of Load for the first [F (1, 53) = 6.55, p = .013, ES = .11] but not for the last stimulus,
363
suggesting our general load effect was driven mostly by the first stimulus. Also, no significant Group-
364
by-Load interaction was found for both first and last stimuli.
365
Our Step-3 analysis, testing for sequence effects of stimuli, revealed a main effect of Stimulus,
366
[F (1, 53) = 11.76, p = .001, ES = .18], and a main effect of Group, [F (1, 53) = 4.89, p = .031, ES = .08],
367
during the low load. No Group-by-Stimulus interaction effect was found. The high load also revealed a
368
main effect of Stimulus, [F (1, 53) = 3.17, p = .05, ES = .11], and a main effect of Group, [F (1, 52) =
369
7.63, p = .008, ES = .13], with no Group-by-Stimulus interaction. In general, for both WM loads,
15
WM encoding in ADHD
370
responses to the first stimulus showed lower amplitudes than to the other stimuli (p‟s < .05). For the
371
high load, pairwise comparisons showed that the second and third stimulus did not differ. The ADHD
372
group showed lower P3 amplitudes for all three stimuli compared to the control group.
373
For the P3 latency, the Step-1 latency analysis indicated no main effect of Group, a marginally
374
significant effect for Load, [F (1, 53) = 3.23, p = .078, ES = .058], with shorter latencies for the high
375
load, and no Load-by-Group interaction. Step-2 analysis revealed a main effect of Load for the first [F
376
(1, 53) = 8.16, p = .006, ES = .133], but not for the last stimulus. No main effect of Group or Group-by-
377
Load interaction was found for either the first or last stimulus. The Step-3 analysis showed a main effect
378
of Stimulus for the low load condition [F (1, 53) = 5.47, p = .023, ES = .093]. Post-hoc tests revealed
379
that the ADHD group showed increased P3 amplitude for the second stimulus compared to the first,
380
whereas the comparison group did not show this difference (p = .03). No main effect of Group or
381
Group-by-Stimulus interaction was found for both first and last stimuli.
382 383
>
384 385
>
386 387
>
388 389 390
DISCUSSION The present study is the first to investigate neural changes in P3 amplitudes and latencies during
391
the encoding phase of WM in ADHD, using a delayed match-to-sample task. The main finding was a
392
reduced P3 amplitude for the ADHD group, regardless of WM load or stimulus sequence (i.e., first or
393
last), compared to the control group. Behavioral results showed that all participants were less accurate
16
WM encoding in ADHD
394
during the high WM load condition compared to the low load condition, which attests to the success of
395
our experimental manipulation of WM load.
396
Changes in P3 amplitudes during WM tasks have been associated with differences in the
397
allocation of attention necessary for WM functioning (Donchin and Coles, 1988; Polich, 2007; Kok,
398
2001). Our findings of lower P3 amplitude in the ADHD group compared to the control group,
399
independent of load or stimulus order, suggests a more general problem of attentional allocation in the
400
encoding phase of WM rather than a more specific problem associated with an increasing demand on
401
WM. The reduced P3 amplitude in college students with ADHD found in this study using a WM task is
402
consistent with and extend findings from a recent meta-analysis on P3 in adults with ADHD (to WM
403
encoding during a WM task), which revealed a decreased P3 amplitude recorded during Go-Nogo tasks
404
in ADHD (Szuromi et al., 2011; and also see, Woltering et al., 2013, using a similar sample to that in
405
the present study). Go-Nogo tasks, which traditionally are used to index response inhibition, do involve
406
WM, albeit low-level demands, and some require constant updating of working memory.
407
We believe that the decreased P3 amplitude observed in individuals with ADHD can be
408
interpreted as deficient attentional allocation, which leads to inefficient encoding of information in
409
memory. Theoretically, the notion of P3 reflecting attentional allocation is not that different to Kok‟s
410
notion of capacity since recent insights in the field of working memory have linked attentional processes
411
and working memory capacity closely together (Awh and Vogel, 2008). A functional relationship
412
between P3 amplitude and attentional resource allocation has also been reported in the literature (Ford,
413
White, Lim, and Pfefferbaum, 1994; Kok, 2001). Also, many studies found the P3 to be related with
414
brain regions thought to be involved in attention, such as the parietal lobe, the temporoparietal junction,
415
lateral prefrontal areas and the cingulate gyrus (Linden, 2005; Verleger et al., 1994). Similarly, a study
416
that explored the cortical basis of ERP components revealed an impaired ventral attentional pathway in
417
ADHD participants during the time of the P3 (Helenius et al., 2011). Furthermore, in their meta-analysis,
418
Szuromi et al (2011) interpreted the overall reduction of P3 amplitude in adults with ADHD as a
17
WM encoding in ADHD
419
dysfunction of the ventral attention network since the P3 related to target detection is thought to be
420
generated from the temporoparietal junction. In sum, previous research suggests a possible link between
421
P3 amplitude and attentional resource allocation. Hence, based on previous findings, we speculate that
422
weak activation in attention pathways might play a role in the defective allocation of attentional
423
resources during activities involving WM, accounting for the observed decreased P3 amplitudes in
424
college students with ADHD.
425
Our analyses of P1 amplitude revealed no significant group effects except for one: namely that
426
the ADHD group seems to manifest an increase in P1 amplitude from the low to the high WM load
427
condition in response to the first stimulus only. Compared to Haenschel et al. (2007), who did find
428
robust P1 group differences in Schizophrenia patients, our findings suggest that differences in WM
429
encoding between the ADHD and non-ADHD peer group were due primarily to differences in
430
attentional allocation during the WM encoding phase as opposed to early visual processing.
431
Only one other study to date has investigated neural differences in WM between ADHD adults
432
and their peers using EEG (Missonnier et al., 2013). Despite the use of different methodologies (n-back
433
versus delayed-match-to-sample task, respectively; smaller sample size, and the focus on different
434
neurophysiological measures, such as time-frequency domain), conclusions are similar. That is, both
435
conclude that individuals with ADHD manifest differences in neural correlates linked to the attentional
436
network involved in the allocation of attention subserving WM.
437
Behavioral differences between groups were found during the WM task: namely the ADHD
438
group had slower response times to the retrieval cue, albeit marginally significant, but did not differ in
439
response accuracy. It is possible that the slower response times in the ADHD group indicate their need
440
for more time to make accurate decisions compared to their peers. In this sense, the longer response
441
times may be another manifestation of slow processing speed or a compensatory mechanism. Although
442
accuracy in our ADHD group was comparable to that of their typically developing peers, cognitive
443
mechanisms of processing, such as the allocation of attention, could still function less optimally. It is
18
WM encoding in ADHD
444
possible that this deficit may show in contexts other than within the experimental confines of the current
445
task. This suggestion is strengthened by other findings from this study: namely that for the ADHD
446
group, their mean scores for WAIS Digit Span and CANTAB SWM were in low average range, whereas
447
those of the control group were in the average range; the T-scores for the ADHD group (mean = 41.5)
448
were in the mildly impaired range whereas those for the control group (mean = 50) were within the
449
well-functioning range. Note that the two neuropsychological tasks not only require the maintenance but
450
also the updating of the spatial representations, which renders their WM demands similar to those of the
451
delayed-match-to-sample task. Furthermore, these behavioral and neuropsychological differences were
452
found despite the fact that the participants with ADHD in this study are high functioning (mostly college
453
students) and the task was self-paced to exclude potential effects of poor sustained attention
454
confounding the WM result.
455
In light of our neural results, the lack of differences in behavioral accuracy in the delayed-
456
match-to-sample task could suggest that our group differences found in neural processing indicate
457
inefficient means of attentional allocation; a detriment that may not always be observed in behavioural
458
accuracy. The P3 could be a neural indicator of this inefficient mechanism.
459
From a clinical perspective, the WM impairment (albeit modest) in these college students with
460
ADHD compared to their college-peers may well help explain why this relatively successful subgroup
461
continues to manifest functional impairments in academic, social functioning during college years.
462
Furthermore, these results support findings from a recent meta-analysis suggesting long-term memory
463
difficulties in adult ADHD may be attributable to difficulties in the encoding phase (Skodzik et al.,
464
2013). Our findings highlight the importance of studying WM encoding and in this newly emergent
465
subgroup of young adults with ADHD.
466
This study has several limitations, which should be taken into consideration when interpreting
467
the findings. First of all, we did not have a direct measure of IQ, or scholastic ability, which limits the
468
characterization of our sample. However, these college students must be of at least average intelligence
19
WM encoding in ADHD
469
(as suggested by the WAIS Digit Span) to be able to successfully complete high school and gain college
470
entrance. Furthermore, we did not measure alcohol and recreational drug use in this sample of college
471
students, which could influence allocation of attention or WM or both. Also, the current analyses were
472
restricted to the encoding stage and further research is warranted to characterize the maintenance and
473
retrieval stages of WM.
474
In conclusion, keeping in mind the aforementioned limitations, our results suggest differences in
475
neural processing in WM with adult ADHD. Our results may suggest, based on what is known about
476
changes in P3 amplitudes in the ERP literature, that weak activation in attentional pathways might play
477
a role in defective attention resource allocation to updating representations in working memory.
478
Moreover, our results concerning the P1 versus P3 suggest a temporal as well as syndrome specific
479
neural index of adult ADHD. These findings could have important implications for understanding the
480
cognitive difficulties ADHD adults face, as well as intervention.
481 482 483
20
WM encoding in ADHD
484
Acknowledgements
485
We thank Dr. Corinna Haenschel and Niko Kriegeskorte for letting us use the task stimuli. Also, we
486
appreciate Alex Lamey for his programming assistance with the E-prime task and Peter Fettes for
487
helping with parts of the data analysis. We would like to thank Dr. Marc Lewis for the use of his
488
Canadian Foundation for Innovation (CFI #482246) funded EEG lab. This research was supported
489
financially in part by a CIHR Operating grant (# 245899, Tannock & Lewis) and by the Canada
490
Research Chair program (RT).
491 492
21
WM encoding in ADHD
493
References
494
Adler LA, Spencer T, Faraone SV, Kessler RC, Howes MJ, Biederman J, et al. Validity of pilot Adult
495
ADHD Self- Report Scale (ASRS) to Rate Adult ADHD symptoms. Ann Clin Psychiatry
496
2006;18:145-148.
497
Awh E, Vogel EK. The bouncer in the brain. Nat Neurosci 2008; 11:5–6.
498
Baddeley A. Working memory: Looking back and looking forward. Nat Rev Neurosci 2003; 4:829-839.
499
Baddeley AD, Hitch GJ. Developments in the concept of working memory. Neuropsychology
500 501 502
1994; 8:485. Barkley RA, Cook E, Diamond A, Zametkin A, Thapar A, Teeter A. International consensus statement on ADHD. Clin Child Fam Psych 2002; 5:89-111.
503
Carrasco M. Visual attention. Vision Res 2011; 51:1484-1525.
504
Cowan N. The magical number 4 in short-term memory: A reconsideration of mental storage capacity.
505 506 507 508 509 510 511
Behav Brain Sci 2001; 24:87-114. Donchin E, Coles MG. Is the P300 component a manifestation of context updating? Behav Brain Sci1988:11:357-374. DuPaul G, Weyandt L, O‟Dell S, Varejao M. College students with ADHD: Current Status and future directions. J Atten Disord 2009; 13:234. Fabiani M, Karis D, Donchin E. P300 and recall in an incidental memory paradigm. Psychophysiology 1986; 23:298-308.
512
Faraone SV, Biederman J, Mick E. The age-dependent decline of attention deficit hyperactivity
513
disorder: a meta-analysis of follow-up studies. Psychol Med 2006; 36:159-166.
514 515 516 517
Ford JM, White P, Lim KO, Pfefferbaum A. Schizophrenics have fewer and smaller P300s: A singletrial analysis. Biol Psychiat 1994; 35:96-103. Frazier TW, Youngstrom EA, Glutting JJ, Watkins MW. ADHD and achievement meta-analysis of the child, adolescent and adult literatures and a concomitant study with college students. J Learn
22
WM encoding in ADHD
518 519
Disabil 2007; 40:49-65. Haenschel C, Bittner RA, Haertling F, Rotarska-Jagiela A, Maurer K, Singer W, et al. Contribution of
520
impaired early-stage visual processing to working memory dysfunction in adolescents with
521
schizophrenia: a study with event-related potentials and functional magnetic resonance imaging.
522
Arch Gen Psychiat 2007; 64:1229.
523
Heiligenstein E, Guenther G, Levy A, Sevino F, Fulwiler J. Psychological and academic
524
functioning in college students with attention deficit hyperactivity disorder. J Am Coll Health
525
1999; 47:181-185.
526 527 528 529 530 531 532 533 534 535
Helenius P, Laasonen M, Hokkanen L, Paetau R, Niemivirta M. Impaired engagement of the ventral attentional pathway in ADHD. Neuropsychologia 2011; 49:1889-1896. Hervey AS, Epstein JN, Curry JF. Neuropsychology of adults with attention-Deficit/hyperactivity disorder: a meta-analytic review. Neuropsychology 2004; 18:485. Karis D, Fabiani M, Donchin E. “P300” and memory: Individual differences in the von Restorff effect. Cognitive Psychol 1984; 16:177-216. Kane MJ, Conway AR, Hambrick DZ, Engle RW. Variation in working memory capacity as variation in executive attention and control. Oxford: Oxford University Press, 2007. Kok A. On the utility of P3 amplitude as a measure of processing capacity. Psychophysiology 2001; 38:557-577.
536
Kok A, Ramautar JR, De Ruiter MB, Band GP, Ridderinkhof KR. ERP components associated with
537
successful and unsuccessful stopping in a stop‐signal task. Psychophysiology 2004; 41:9-20.
538 539 540 541 542
Linden DEJ. The P300: where in the brain is it produced and what does it tell us? Neuroscientist 2005; 11:563–76. Luck SJ. An introduction to the event-related potential technique (cognitive neuroscience). Cambridge, MA: MIT Press, 2005.
Martinussen R, Hayden J, Hogg-Johnson S, Tannock R. A meta-analysis of
23
WM encoding in ADHD
543
working memory impairments in children with attention-deficit/hyperactivity disorder. J
544
Am Acad Child Adolesc Psychiatry 2005;44:377-384.
545
Missonnier P, Hasler R, Perroud N, Herrmann FR, Millet P, Richiardi J, et al. EEG anomalies in adult
546
ADHD participants performing a working memory task. Neuroscience 2013; 241:135-146.
547
Polanczyk G, Jensen P. Epidemiologic considerations in attention deficit hyperactivity disorder: a
548 549 550
review and update. Child Adol Psych Cl 2008; 17:245-260. Polanczyk G, Rohde LA. Epidemiology of attention-deficit/hyperactivity disorder across the lifespan. Curr Opin Psychiatry 2007;20:386-389.
551
Polich J. Updating P300: an integrative theory of P3a and P3b. Clin Neurophysiol 2007; 118:2128-2148.
552
Rutman AM, Clapp WC, Chadick JZ, Gazzaley A. Early top–down control of visual processing predicts
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
working memory performance. J Cognitive Neurosci 2010; 22:1224-1234. Shaw-Zirt B, Popali-Lehane L, Chaplin W, Bergman A. Adjustment, social skills, and self-esteem in college students with symptoms of ADHD. J Atten Disord 2005; 8:109-120. Skodzik T, Holling H, Pedersen A. Long-Term Memory Performance in Adult ADHD: A Meta-Analysis. J Atten Disord 2013; doi 10.1177/1087054713510561. Szuromi B, Czobor P, Komlosi S, Bitter I. P300 deficits in adults with attention deficit hyperactivity disorder: a meta-analysis. Psychol Med 2011; 41:1529. Tabachnick BG, Fidell LS. Multivariate analysis of variance and covariance using multivariate statistics. Boston: Allyn and Bacon, 2001. Verleger R, Heide W, Butt C, Kömpf D. Reduction of P3b in patients with temporo-parietal lesions. Brain Res 1994; 2:103-116. Vogel EK, McCollough AW, Machizawa MG. Neural measures reveal individual differences in controlling access to working memory. Nature 2005; 438:500-503. Wallace JC, Kass SJ, Stanny CJ. The Cognitive Failures Questionnaire revisited: Dimensions and correlates. J Gen Psychol 2002; 129:238-256.
24
WM encoding in ADHD
568 569 570 571
Woltering S, Granic I, Lamm C, Lewis MD. Neural changes associated with treatment outcome in children with externalizing problems. Biol Psychiat 2011; 70:873-879. Woltering S, Liu Z, Rokeach A, Tannock R. Neurophysiological differences in inhibitory control between adults with ADHD and their peers. Neuropsychologia 2013; 51: 1888-1895.
572 573 574
25
WM encoding in ADHD
575
Figure Legends
576 577
Figure 1.Schematic outline of the delayed match-to-sample task for low load condition.
578
Working memory paradigm adopted and modified from Haenschel et al (2007). Two stimuli were
579
presented in low load condition, and 3 stimuli were shown in high load condition for encoding for 600
580
ms each with an inter-stimulus interval of 400 ms. After a 2000 ms delay, a probe stimulus was
581
presented for the participants to judge whether they saw the target among the initial set of the stimuli or
582
not. Participants initiated the next trial by pressing the space bar. 200 ms of clear time was allowed
583
before each trial.
584 585
Figure 2. ERP waveforms at site P3 (averaged across all P3 electrodes) for (A) Low and (B) High
586
load conditions for each stimulus [(A) left: first stimulus, right: second stimuli (B) first, second
587
and last stimuli].
588 589
Figure 3. ERP waveforms at site P3 (averaged across all P3 electrodes) and topoplots (250-500 ms)
590
for (A) Low and (B) high load condition across the stimuli.
591
26
Figure 1
Encoding (600ms per stimuli) 600ms
600ms
Maintenance (delay: 2s)
Retrieval (target:2s) 2000ms
Figure 2
Figure 3
Table 1: Mean and Standard deviations of the ASRS (ADHD Self Report Scale), CFQ (Cognitive Failures Questionnaire), and SA-45 subscales (Symptom Assessment-45) for the ADHD and the Comparison group.
ADHD
Comparison
P-values
ASRS ***
44.97(9.12)
18.71(7.98)
.00
CFQ ***
53.88(12.74)
27.56(8.09)
.00
Anxiety*
8.41(2.37)
6.75(1.83)
.01
Depression
9.37(3.63)
8.05(3.58)
.21
14.72(3.77)
9.25(3.45)
.00
Somatization
7.65(2.59)
6.55(1.73)
.11
Phobic Anxiety
5.52(0.80)
5.29(0.72)
.30
Hostility
6.48(1.71)
5.95(1.50)
.28
Interpersonal Sensitivity
8.67(3.16)
7.76(3.85)
.38
Paranoid Ideation
7.93(2.56)
6.62(2.20)
.07
Psychoticism
5.54(1.07)
5.62(1.02)
.80
SA-45
Obsessive Compulsive ***
* p < .05, ** p < .01, *** p < .001. p-values shown reflect one-way ANOVA’s.
Table 2. Mean and Standard deviations for behavioural measures of the delayed match to sample task [accuracy, as well as the reaction time (in milliseconds), the intra-individual variability (IIV)], and the WM [Digit span, CANTAB: SWM and SSP] neuropsychological tests for the ADHD and Comparison group. ADHD Mean (SD)
Comparison Mean (SD)
P-values
Delayed Match to Sample Task Accuracy Low load
0.73(0.14)
0.80(0.14)
.10
High load
0.62(0.18)
0.64(0.15)
.64
Low load
1260.60(232.85)
1176.17(166.47)
.14
High Load
1309.96(268.49)
1241.70(208.08)
.30
Low load
494.32(241.75)
446.06(216.30)
.44
High load
420.31(242.63)
452.05(228.98)
.62
23.32(3.26)
21.81(2.57)
.07
WAIS Digit Span
8.31(3.31)
9.95(2.94)
.08
CANTAB SWM*
-0.10(1.13)
0.51(0.62)
.04
0.09(0.86)
0.60(1.14)
.10
Reaction Time
Intra-individual variability (RT)
Total duration of the task Neuropsychological tests
SSP
* p < .05, p-values shown reflect one-way ANOVA’s, RT (Reaction Time), SWM (Spatial Working Memory), SSP (Spatial Span)
Table 3: Means and standard deviations for ERP amplitudes and latency. ADHD Amplitude
P1 Low load 1st stimulus 2nd stimulus High load 1st stimulus 2nd stimulus 3rd stimulus P3 Low load 1st stimulus 2nd stimulus High load 1st stimulus 2nd stimulus 3rd stimulus
Latency
Control
P1 Low load 1st stimulus 2nd stimulus High load 1st stimulus 2nd stimulus 3rd stimulus P3 Low load 1st stimulus 2nd stimulus High load 1st stimulus 2nd stimulus 3rd stimulus
3.27(1.30) 2.83(1.51) 3.70(1.54) 3.62(1.22) 3.20(1.52) 3.87(1.70) 3.83(1.13)
4.02(2.09) 3.83(2.35) 4.21(2.05) 4.30(2.08) 3.58(2.42) 4.33(1.90) 5.05(2.36)
1.96(1.52) 1.59(1.70) 2.32(1.68) 2.02(1.45) 1.80(1.42) 2.18(1.60) 2.04(1.61)
3.02(2.06) 2.69(2.61) 3.39(1.85) 3.32(2.07) 3.12(2.20) 3.47(1.99) 3.29(2.18)
127.27(16.98) 127.27(16.98) 124.47(16.19) 125.58(11.99) 124.73(15.21) 124.53(13.47) 126.55(14.19)
127.92(17.05 127.92(17.05) 124.80(16.30) 125.65(11.66) 127.76(19.41) 123.44(16.16) 125.76(10.30)
360.13(26.42) 350.27(42.41) 370.00(38.08) 468.04(29.31) 372.60(48.80) 366.27(35.64) 365.27(42.75)
357.40(22.59) 350.24(37.86) 364.56(23.31) 362.21(18.94) 363.52(31.73) 360.72(24.66) 362.40(25.23)