You are on page 1of 7

Intelligence 40 (2012) 100–106

Contents lists available at SciVerse ScienceDirect

Intelligence

Bottom–up mechanisms are involved in the relation between accuracy in


timing tasks and intelligence — Further evidence using manipulations of
state motivation
Fredrik Ullén a,⁎, Therese Söderlund b, Lenita Kääriä b, Guy Madison b
a
Dept. of Neuroscience, Karolinska Institutet, SE-171 77 Stockholm, Sweden
b
Dept. of Psychology, Umeå University, SE-901 87 Umeå, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: Intelligence correlates with accuracy in various timing tasks. Such correlations could be due to
Received 29 October 2009 both bottom–up mechanisms, e.g. neural properties that influence both temporal accuracy and
Received in revised form 24 January 2012 cognitive processing, and differences in top–down control. We have investigated the timing–
Accepted 24 January 2012 intelligence relation using a simple temporal motor task, isochronous serial interval produc-
Available online 13 February 2012
tion (ISIP), i.e. hand/finger movements with a regular beat. ISIP variability is negatively corre-
lated with intelligence and we have previously argued, based on indirect evidence, that this
Keywords: relation has a bottom–up component. Here, we investigate this question using an experimen-
Timing tal within-subject design in two samples (n = 38 and n = 95 participants, respectively). ISIP
Tapping
was performed under two conditions. In the first condition (Low Motivation), the participants
Incentive
were told that measurements were being made to familiarize them with the task and to cali-
Elementary cognitive tasks
brate the equipment. In the second condition (High Motivation), the participants were told
that the performance would be evaluated and used for scientific analysis, and they were
given a monetary reward depending on how accurately they performed. Temporal accuracy
in the ISIP was higher during High Motivation than during Low Motivation. In both samples,
correlations between ISIP variability and intelligence were similar for both conditions. General
linear models with ISIP variability measures as dependent variables, condition (Low Motiva-
tion or High Motivation) as a repeated-measures variable and intelligence as a between-
subject variable, revealed a significant effect of intelligence, but no effects of incentive, nor
of the intelligence × incentive interaction. We conclude that motivationally driven top–down
mechanisms can influence ISIP performance, but that they play no major role for correlations
between temporal accuracy in ISIP and intelligence. These results provide further support for
that bottom–up mechanisms are involved in relations between temporal accuracy and
intelligence.
© 2012 Elsevier Inc. All rights reserved.

1. Introduction Karabanov, & Madison, 2008), but why is this the case? In
principle, such correlations could involve both bottom–up
Intelligence correlates with accuracy in various tasks that and top–down mechanisms. This distinction is theoretically
involve processing of temporal information (Helmbold, important. If bottom–up mechanisms are involved, it sug-
Troche, & Rammsayer, 2007; Madison, Forsman, Blom, gests that timing tasks tap basic neural properties that also
Karabanov, & Ullén, 2009; Rammsayer & Brandler, 2002, play a role for cognition. In this case, temporal accuracy of
2007; Troche & Rammsayer, 2009; Ullén, Forsman, Blom, neural activity may be a causal factor underlying differences
in intelligence. Top–down mechanisms such as attention, on
⁎ Corresponding author. the other hand, are likely to be general and influence perfor-
E-mail address: Fredrik.Ullen@ki.se (F. Ullén). mance in both temporal and non-temporal tasks. If the

0160-2896/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.intell.2012.01.012
F. Ullén et al. / Intelligence 40 (2012) 100–106 101

relations between intelligence and temporal tasks depend portion of the ISIP–intelligence correlations, in particular
solely on differences in top–down control, temporal accuracy with regard to drift. On the other hand, as argued above, sev-
may play no specific roles for cognitive performance. eral findings on ISIP–intelligence relations, as well as the fact
A number of recent studies have shed light on the nature that temporal discrimination ability predicts intelligence
of intelligence–timing relations. Rammsayer and coworkers over and above general discrimination ability, suggest that
have demonstrated that performance in temporal judgment bottom–up mechanisms are involved as well.
and discrimination tasks that involve maintenance and ma- Here, we used a within-subject design to investigate the
nipulation of temporal information in working memory cor- nature of the ISIP–intelligence relation further. The objective
relates with intelligence (Helmbold et al., 2007; Rammsayer was to study the effects of manipulating incentive motivation
& Brandler, 2002, 2007). Recent findings from this group during ISIP performance on intelligence–ISIP relations. ISIP
show that intelligence also correlates with discrimination of was performed under two conditions. In the first condition
non-temporal information (pitch, loudness, brightness) and, (Low Motivation), the participants were instructed that mea-
furthermore, that temporal and non-temporal discrimination surements were being made to familiarize them with the task
abilities are substantially intercorrelated and may reflect a and to calibrate the equipment. In the second condition (High
general discrimination ability that, in turn, is related to intel- Motivation), the participants were told that the performance
ligence (Troche & Rammsayer, 2009). Interestingly, latent would be evaluated and used for scientific analysis, and that
variable models showed that temporal discrimination ability they would be given a monetary reward depending on how
predicts speed-related aspects of psychometric intelligence accurately they performed.
independently of this general discrimination ability. The rationale for the design was as follows. State motiva-
We (Forsman, Madison, & Ullén, 2009; Holm, Ullén, & tion can influence performance by recruiting neural circuitry
Madison, 2011; Madison et al., 2009; Ullén et al., 2008) involved in attentional control to increase selective attention
have studied intelligence–timing relations using a simple to task (Pessoa, 2009). If individual differences in such top–
motor timing task, isochronous serial interval production down mechanisms are important for the intelligence–ISIP re-
(ISIP). In this task the participant produces a series of repeti- lation, we would predict higher correlations during the High
tive rhythmic hand or finger movements with a regular beat. Motivation condition, when these mechanisms are recruited,
The total variability of the produced temporal intervals can than during Low Motivation. If, on the other hand, top–down
be divided into local variability (Local) between neighboring mechanisms recruited by motivation are of minor impor-
intervals, and drift (Drift), i.e. fluctuations in tempo over sev- tance for the intelligence–ISIP relation we would expect the
eral intervals. We have previously argued that Local variabil- same slope of this relation in both conditions. We tested
ity in ISIP is, at least in part, automatic and inaccessible to these predictions in two samples, using general linear models
conscious control, whereas Drift may be dependent on with Local or Drift as dependent variables, condition (High or
short-term memory of previously produced intervals and Low Motivation) as a categorical within-subject factor and
thus more susceptible to top–down influences (Forsman et intelligence as a continuous between-subject factor. The
al., 2009; Madison & Delignières, 2009; Madison et al., 2009). presence of condition × intelligence interactions in these
Several observations suggest that bottom–up mechanisms models would suggest that motivationally controlled top–
are involved in the ISIP–intelligence relation. Firstly, both down mechanisms are important for relations between ISIP
Local and Drift correlate negatively with intelligence and intelligence. Sample 1 was a smaller student sample
(Madison et al., 2009; Ullén et al., 2008). Secondly, correla- where intelligence was measured using the Wiener Matrizen
tions between intelligence and ISIP are strongest for tapping Test. Findings from this sample were replicated in a larger
intervals in the range 0.5–1 s (Madison et al., 2009). A consis- second sample, where the Raven SPM Plus was used as an in-
tent picture from the neuroimaging literature is that motor telligence measure.
timing in the subsecond range loads little on brain regions in-
volved in cognitive control (Lewis & Miall, 2003). Thirdly, no 2. Methods
worst performance rule could be found for the ISIP–intelli-
gence relation (Madison et al., 2009). The worst performance 2.1. Participants
rule has been observed in studies of reaction times in ele-
mentary cognitive tasks and their correlation with intelli- 2.1.1. Sample 1
gence. It implies that the longest reaction times produced The participants were students from Umeå University
by an individual predict intelligence better than do the short- (Umeå, Sweden) who took part in the study as part of their
est reaction times (Coyle, 2003). This phenomenon presum- psychology education or were recruited via advertisements
ably reflects that the longest reaction times occur during on bulletin boards at the campus. Thirty-eight individuals
attentional lapses, which are correlated with intelligence agreed to participate. Of these, three had to be excluded due
(Coyle, 2003; Schweizer & Moosbrugger, 2004; Unsworth, to erratic performance in the rhythmic tasks. Thirty-five partic-
Redick, Lakey, & Young, 2010). The absence of a worst perfor- ipants (18 males) aged 21 to 31 years [mean (M) = 24.9; stan-
mance rule for ISIP suggests that lapses of attention are of dard deviation (SD)= 2.8] were thus included in the analyses.
minor importance for ISIP–intelligence correlations.
A reasonable hypothesis based on these findings, taken 2.1.2. Sample 2
together, is thus that timing–intelligence relations involve These participants were recruited through posters on
both top–down and bottom–up mechanisms. Individual dif- public bulletin boards in Stockholm and Umeå (Sweden). Of
ferences in general top–down control are likely to be in- the 98 participants that agreed to participate, two were ex-
volved in general discrimination ability as well as in a cluded due to erratic performance in the rhythmic tasks,
102 F. Ullén et al. / Intelligence 40 (2012) 100–106

and one due to failure to complete the intelligence test. Only the shorter trials (30 beats) were used in the present
Ninety-five participants (39 males) were thus included in study. (The longer trials were collected for an unrelated
the analyses. Their age ranged from 18 to 49 years study on long-range temporal variability.) This procedure
(M = 27.5; SD = 6.5). For two of these participants data was repeated for each trial, which had one of four different
were missing in one of the two conditions due to technical stimulus inter-onset-intervals (IOIs) (524, 655, 1024, and
problems with the recording equipment. 1624 ms). Only the three shorter IOIs were used in the pre-
The experimental procedures were undertaken with the sent analyses, since earlier studies found that correlations be-
understanding and written consent of each participant and tween intelligence and ISIP variability are weak for IOIs above
conformed to The Code of Ethics of the World Medical Asso- 1 s (Madison et al., 2009).
ciation (Declaration of Helsinki). Ethical approval was given The experiment started with three practice trials with the
by the Ethical Committee of Umeå University. IOIs 500, 733, and 1044 ms. This was followed by two exper-
imental blocks with eight trials each, first all IOI with 30 re-
2.2. Materials sponses and then the same IOIs again with 145 responses.
IOIs were always presented in the same ascending order.
The ISIP task was implemented in custom designed soft- Task instructions differed between the two experimental
ware running on a PC with a real-time operating system. An blocks. Before one of the blocks, the High Motivation condi-
Alesis D4 drum module connected via MIDI to the PC pro- tion, the participant was told that the performance was
duced the sounds and collected the responses. The temporal being evaluated and measured and that this was the real ex-
resolution of the system was 1 ms. Stimuli consisted of 20 periment. They were guaranteed a monetary reward of 30
sampled cowbell sounds presented in isochronous sequence Swedish crowns but were told they could receive up to 70
through Peltor HTB7A sound-attenuated headphones at crowns depending on the accuracy of their performance. In
78 dBA sound pressure level. The last two sounds were atten- the Low Motivation condition the participants were told
uated to 72 and 66 dBA, respectively, to reduce the startle re- that the purpose of the recording was partly to familiarize
action when stimuli cease. The sounds had a supra-threshold them with the task but mainly to calibrate the instrument
duration of approximately 80 ms. Responses were given by and obtain baseline reference data. The order of the two con-
beating a drumstick against a drum pad with a piezoelectric ditions was counterbalanced across the participants. Partici-
element. pants with Low Motivation as second condition were given
In Sample 1, intelligence was estimated using the Wiener the same information as participants with Low Motivation
Matrizen Test (WMT). This test is a timed (25 min) 24-item as first condition except that it was not said, for obvious rea-
test which is similar in construction to, and highly correlated sons, that it was a purpose of the condition to familiarize
with, Raven's Progressive Matrices test (Formann & them with the task. In both conditions the participants
Piswanger, 1979). In Sample 2, intelligence was measured were told to read the information appearing on the computer
using the Raven SPM Plus (Raven, Raven, & Court, 2000), an screen after each sequence. In the Low Motivation condition
untimed 60-item version of the Raven's Progressive Matrices the information consisted of a sentence confirming that the
test that is commonly used as a measure of general fluid in- sequence had been registered. In the High Motivation condi-
telligence (Gustafsson, 1984; Jensen, 1998). tion the information was false evaluative performance feed-
A questionnaire with three visual analog scale items was back intended to be motivating. The feedback functioned as
used to measure subjective motivation during the High Moti- a goal-setting promoter and was always the same for every
vation and Low Motivation conditions of the ISIP task, and participant.
the WMT. The participant rated motivation by indicating a After the ISIP task the participants were offered a break of
cross on horizontal line that ranged from “not at all motivat- a few minutes. They were then subjected to the WMT, which
ed” (left endpoint) to “very motivated” (right endpoint). was administered as prescribed in the manual, i.e. with a
During scoring the scale was divided into 10 equal steps so 25 minute time limit (Formann & Piswanger, 1979). After
that scores on each item ranged between 0 and 10. The par- this, the participants filled in the questionnaire on subjective
ticipants were also given questionnaires about creative motivation during the ISIP and WMT tasks. Finally, a number
achievement and flow experiences, and the NEO personality of questionnaires (creative achievement, flow experiences,
questionnaire. Data from these tests are not used in the pre- NEO personality) that were not used in the present study
sent study. Statistical analyses were performed in Statistica were administered.
8.0 (StatSoft, Inc.).
2.4. Dependent measure computation
2.3. Procedure
The raw number of correct items on the intelligence test
The ISIP task was performed identically in Samples 1 and was used as a measure of intelligence. Processing and analy-
2. Each participant was tested individually, sitting upright sis of the ISIP data followed the procedures developed and
on a chair with his or her feet to the floor. The ISIP task was described previously (Madison, 2001; Madison et al., 2009;
performed first. As described previously (Madison et al., Ullén et al., 2008). In brief, only the data from the continua-
2009), each ISIP trial consisted of an initial synchronization tion phases of the experimental trials were used, excluding
phase, where the participant synchronized right hand move- the first five data points of each trial. After exclusion of outli-
ments with 20 stimulus sounds, followed by a continuation er intervals, the total variance in ISIP data was partitioned
phase where the participant continued to beat another 30 or into two components. Local variability (Local) was estimated
145 times without interruption after the sounds had stopped. as the variance of difference scores between temporal
F. Ullén et al. / Intelligence 40 (2012) 100–106 103

intervals two intervals apart, i.e. with a lag 2 difference. This Table 1
estimate is minimally influenced by gradual changes in the ISIP performance under Low and High Motivation.

duration of the produced intervals, i.e. drift. Drift was esti- Low High Repeated measure ANOVA
mated as the remaining portion of the variance, i.e. total var- Motivation Motivation
iance − local variance, which consists mainly of the
Mean SD Mean SD F value p value partial η2
aforementioned drift in the mean interval. Local and Drift
Sample 1
measures for each trial were divided with the mean duration
Local 3.86 1.01 3.52 .88 11.87 .00077 .26
of all intervals in a trial, to make the measures comparable to Drift 2.95 2.15 2.41 1.47 2.94 .048 .080
coefficients of variation. Analyses were based on the within- Sample 2
participant mean values of Local and Drift for all ISIP trials Local 3.80 1.19 3.61 1.24 3.28 .037 .034
with IOIs 524, 655 and 1024 ms (see above). Drift 1.64 1.92 1.46 1.23 1.36 .12 .015

2.5. Statistical analyses


Table 2 summarizes raw correlations between intelli-
The normality of the distribution of intelligence test gence and ISIP performance. In Sample 1, intelligence was
scores was tested using the Shapiro–Wilk W test. Pearson measured with the WMT test. WMT scores ranged between
product-moment correlations were used to examine rela- 8 and 23 (M = 16.0, SD = 3.9). The distribution of scores
tions between ISIP performance and intelligence for descrip- was not significantly different from normal (Shapiro–Wilk
tive purposes. Effects of incentive motivation on ISIP W = .96, p = .30). Correlations between ISIP performance
performance were calculated using repeated measure ANO- and WMT scores were significant and highly similar in the
VAs. Local, Drift or scores on the motivation questionnaire two conditions ranging from r = −.40 to r = −.36 (Table 2).
were used as dependent variable and motivation as a In Sample 2, the Raven SPM Plus was used to measure intel-
within-participant categorical predictor (corresponding to ligence. SPM scores ranged from 31 to 60 (M = 46.4,
the High Motivation or Low Motivation conditions). General SD = 6.43), and the distribution did not differ from the nor-
linear models were used to test whether the degree of im- mal (Shapiro–Wilk W = .99, p = .63). Significant correlations
provement of performance with incentive was related to in- between ISIP performance and SPM scores were found, rang-
telligence, i.e. whether there was a difference in slope in the ing from r = −.25 to r = −.22 (Table 2).
intelligence–variability relation depending on motivation. We next tested for significant effects of condition on the
Here, Local or Drift was used dependent variable, condition slope of the intelligence × ISIP relation. For this purpose, we
(High Motivation or Low Motivation) as within-subject cate- used general linear models with either Local or Drift as de-
gorical factor and intelligence as between-subject indepen- pendent variable, condition as categorical within-subject
dent variable. factor (High Motivation or Low Motivation) and intelligence
To test for the importance of task order effects, two addi- as a between-subject variable. In Sample 1, with Local as de-
tional analyses were performed in the larger Sample 2. First, pendent variable, there was a significant effect of intelli-
general linear models where task-order and its interactions gence [F(1, 33) = 6.9, p = .01, partial η 2 = .17], but no
with the other independent variables (i.e. task order × moti- effects of condition [F(1, 33) = 1.7, p = .2, partial η 2 = .049]
vation, task order × intelligence, and task order × motivatio- or the interaction condition × intelligence [F(1, 33) = .27,
n × intelligence) had been added to the models discussed p = .60, partial η 2 = .0082]. Similarly, when using Drift as a
above were inspected. The purpose of this analysis was to dependent variable, there was a significant effect of intelli-
see whether task order influenced any motivational influ- gence [F(1,33) = 7.6, p = .01, partial η 2 = .19], but no effects
ences on the ISIP–intelligence relation. Secondly, a of condition [F(1, 33) = 1.75, p = .19, partial η 2 = .050] or of
between-group analysis was performed, where Sample 2 condition × intelligence [F(1, 33) = .91, p = .35, partial
was split into two subsamples depending on task order. Sub- η 2 = .027]. The same pattern of effects was found in Sample
sample 2a thus started with Low Motivation, whereas Sub- 2. With Local as dependent variable, there was a significant
sample 2b started with High Motivation. The effect of intelligence [F(1, 91) = 7.34, p = .008, partial
intelligence × Local and intelligence × Drift correlations dur- η 2 = .075] and no effects of condition [F(1, 91) = .0002,
ing Low Motivation in Subsample 2a were compared with p = .99, partial η 2 b .00001] or condition × intelligence [F(1,
the corresponding correlations during High Motivation in 91) = .056, p = .81, partial η 2 = .00061]. Also for Drift, the ef-
Subsample 2b, using a Difference Test for Pearson r:s. In this fect of intelligence was significant [F(1, 91) = 6.51, p = .012,
analysis, any carry over effects from a previously performed
condition are excluded, since only the first condition per-
Table 2
formed by each participant is considered. Correlations between intelligence and ISIP variability. Pearson r values and
their corresponding p values are provided.
3. Results
Local Drift

ISIP performance in the two samples is summarized in Low High Low High
Table 1. Repeated measure ANOVAs revealed that perfor- Motivation Motivation Motivation Motivation

mance was significantly more accurate during High Motiva- Sample 1 −.40 −.39 −.39 −.36
tion than Low Motivation in both samples. This was seen (p = .02) (p = .02) (p = .02) (p = .03)
Sample 2 −.24 −.25 −.24 −.22
for Local as well as Drift, with the exception for Drift in Sam-
(p = .02) (p = .01) (p = .02) (p = .04)
ple 2.
104 F. Ullén et al. / Intelligence 40 (2012) 100–106

partial η 2 = .067], while the effects of condition [F(1, 91) = finding from earlier investigations (Holm et al., 2011;
1.88, p = .17, partial η 2 = .020] and condition × intelligence Madison et al., 2009; Ullén et al., 2008) that both Local and
[F(1, 91) = 1.50, p = .22, partial η 2 = .016] were not. Drift variability in ISIP are negatively related to intelligence.
To further corroborate that these findings were not con- A summary of all reported relations between intelligence
founded with task order effects two additional analyses was and the ISIP is provided in Table 3. We have earlier observed
performed in the larger Sample 2. The order between Low that a somewhat stronger relation to intelligence is seen for
Motivation and High Motivation was counterbalanced be- Local than for Drift, across tapping frequencies (Madison et
tween subjects. First, we repeated the analyses above for al., 2009). Overall, this trend holds true also for mean values
Sample 2, including task order as a categorical independent across tapping frequencies (Table 3), but the effect size for
variable in the model. In this extended model there was, for both Local and Drift appears to be around r = −.30.
Local, still an effect of intelligence [F(1, 89) = 9.75, p = .002, Increasing motivation improved performance in the ISIP.
partial η 2 = .099] while all effects of condition, task order, This improvement was seen both in Local, i.e. variability be-
motivation and their interactions remained non-significant. tween neighboring intervals, and Drift, i.e. more gradual fluc-
The same pattern was seen for Drift, i.e. the only significant tuations in beat period that involve several intervals. The new
effect was that of intelligence [F(1, 89) = 8.22, p = .005, par- key finding in the study is that these effects on ISIP perfor-
tial η 2 = .085]. mance did not significantly affect relations between ISIP var-
Secondly, we split Sample 2 into two subsamples using iability and intelligence. There was no statistical difference in
task order as a grouping variable. Subsample 2a (n = 46) the slope of the relations between ISIP variability and intelli-
thus started with Low Motivation, while Subsample 2b gence between the Low Motivation and High Motivation con-
(n = 47) started with High Motivation. This allowed us to dition, i.e. the interaction between intelligence and
use a between-group design and compare the ISIP × intelli- motivation had no effect on Local or Drift. In other words,
gence relation during Low Motivation in Subsample 2a, the magnitude of the improvement in ISIP performance
which performed Low Motivation first, with the relation with motivation was unrelated to intelligence. The analyses
found during High Motivation in Subsample 2b, which per- performed in Sample 2 further support that these effects
formed High Motivation first. By only analyzing the first con- are not due to task order confounds, although trends for
dition performed by each participant, carry over effects from higher associations were found for High Motivation.
a previously performed condition are excluded. For Local, the These results are important, in that they provide direct
relation was r = −.09 in Subsample 2a and r = −.29 in Sub- evidence for that those top–down mechanisms that were
sample 2b. This difference in r values was not significant recruited by increased motivation are not essential for ISIP–
(p = .33, Difference Test for Pearson r:s). For Drift, the rela- intelligence relations. This provides further support for that
tion was r = −.13 in Subsample 2a and r = −.30 in Subsam- the relations between temporal variability and intelligence
ple 2b. This difference in r values was significant (p = .41). are not only due to individual differences in general top–
down mechanisms that are employed in both temporal and
4. Discussion non-temporal tasks. Earlier evidence for a bottom–up com-
ponent in relations between temporal variability and intelli-
4.1. Correlations between temporal variability in the ISIP and gence were more indirect: (i) both Local and Drift
intelligence variability in ISIP correlate with intelligence; (ii) these corre-
lations are strongest for interresponse intervals below 1 s,
This study is an investigation of how relations between which is typically considered to be below the threshold for
temporal variability in the ISIP task and intelligence depend cognitive timing; and (iii) ISIP–intelligence relations show
on state motivation during ISIP. In the Sample 1, intelligence no worst performance rule, i.e. the most variable ISIP trials
was measured with the Wiener Matrizen Test, while the do not predict intelligence better than the most stable ones.
Raven SPM Plus was used in Sample 2. Since the experimen- The latter result is particularly relevant for the present
tal procedures otherwise were identical and since findings findings. For reaction time in elementary cognitive tasks a
from the two samples were qualitatively similar they will worst performance rule is typically found, i.e. the longest re-
be discussed in conjunction. In the ISIP task we employed action times correlate stronger with intelligence than do the
interresponse intervals in the range 0.5–1 s, which is the shortest reaction times (Coyle, 2003). This presumably re-
range where previous studies found the strongest relations flects that the longest reaction times occur during lapses in
to intelligence (Madison et al., 2009). We thus replicate the sustained attention, which in turn are correlated with intelli-

Table 3
Effect sizes of reported relations between ISIP and intelligence. For the present study, the mean of the r values found for High Motivation and Low Motivation are
given.

Study Sample Intelligence test Number of participants Intelligence × Local (r value) Intelligence × Drift (r value)

Madison et al. (2009) 1 Raven SPM Plus 34 −.44 −.41


2 Raven SPM Plus 30 −.44 −.36
Holm et al. (2011) 1 Raven SPM Plus 112 −.24 −.19
Present study 1 Wiener Matrizen Test 35 −.40 −.38
2 Raven SPM Plus 95 −.25 −.23
Weighted mean −.30 −.27
F. Ullén et al. / Intelligence 40 (2012) 100–106 105

gence (Unsworth et al., 2010). In contrast, it appears that cor- We have argued earlier that these specific contributions of
relations between ISIP variability and intelligence depend lit- temporal variability to intelligence may represent bottom–up
tle on changes in state attention, whether these are due to mechanisms, i.e. basic neural properties that influence both
spontaneous lapses or — as in the present study — motiva- timing accuracy and intelligence. Previous findings from the
tionally driven. group suggest that the volume of white matter connections
in prefrontal brain regions is one common neural substrate
4.2. Top–down and bottom–up mechanisms in relations be- for intelligence and timing (Ullén et al., 2008). In line with
tween timing tasks and intelligence Jung's and Haier's parieto-frontal model of intelligence
(Jung & Haier, 2007), prefrontal connectivity may thus influ-
In general, individual differences in the function of atten- ence variability of neuronal activity in fronto-parietal circuits
tional systems show substantial correlations with intelli- and this, in turn, is important for intelligence. Clearly, further
gence (Schweizer & Moosbrugger, 2004). A priori, one work is needed to provide an understanding of how variabil-
would therefore predict that attentional mechanisms are ity affects intellectual function. In general terms, it is conceiv-
also responsible for a portion of the correlations between able that millisecond level precision of neuronal discharges
performance in timing tasks and intelligence, in particular affects both neural synchronization, which may be a mecha-
for tasks that require cognitive control, e.g. discriminations nism to dynamically represent both sensory percepts and
or judgments on temporal intervals maintained in working cognitive contents (see e.g. Singer, 1999; Uhlhaas & Singer,
memory. Indeed, as summarized in the Introduction, recent 2006), and synaptic plasticity (Kampa, Letzkus, & Stuart,
results from the group of Rammsayer (Troche & 2007).
Rammsayer, 2009) are in line with this prediction. Temporal
discrimination ability was found to be highly correlated with
a latent general discrimination ability, which influenced dis- Acknowledgments
crimination performance across sensory modalities, and
which in turn was related to intelligence. It appears likely This research was supported by the Swedish Research
that this general discrimination ability in part corresponds Council, the Sven and Dagmar Salén Foundation, the Söder-
to individual differences in broad top–down mechanisms berg Foundation, and the Freemasons in Sweden Foundation
such as the ability to sustain selective attention to one senso- for Children's Welfare. We are thankful to Örjan de Manzano
ry modality or to ignore distracters. However, Troche and for comments on an earlier version of this manuscript.
Rammsayer also found that performance in temporal tasks
predicted intelligence over and above general discrimination
ability. This finding is consistent with the present data on References
ISIP, which suggests that common neural properties affect
temporal variability and cognitive performance indepen- Coyle, T. R. (2003). A review of the worst performance rule: Evidence, theory
and alternative hypotheses. Intelligence, 31(6), 567–587.
dently of attentional control. Deary, I. J. (2000). Looking down on human intelligence: From psychometrics to
We found effects of motivation on both Local and Drift the brain. Oxford: Oxford University Press.
variability, which would seem to indicate that none of them Formann, W., & Piswanger, J. (1979). Wiener Matrizen Test [Vienna Matrices
Test]. Göttingen: Hogrefe Verlag.
tap a pure measure of automatic temporal variability that is Forsman, L., Madison, G., & Ullén, F. (2009). Neuroticism is correlated with
inaccessible to top–down control exclusively. However, we drift in serial time interval production. Personality and Individual Differ-
must caution that Drift is not computationally independent ences, 47(3), 229–232.
Gustafsson, J. -E. (1984). A unifying model for the structure of intellectual
from Local, since they are both based on the same overall var- abilities. Intelligence, 8, 179–203.
iance across a response series. It remains therefore an open Helmbold, N., Troche, S., & Rammsayer, T. (2007). Processing of temporal
question whether their dependency is due to an intrinsic re- and nontemporal information as predictors of psychometric intelli-
gence: A structural-equation-modeling approach. Journal of Personality,
lation between the underlying subprocesses or to the way
75(5), 985–1006.
they are computed. Further development of ISIP variability Holm, L., Ullén, F., & Madison, G. (2011). Intelligence and temporal accuracy
measures, e.g. measures that are derived from models of of behavior: Unique and shared associations with reaction time and
ISIP performance, may allow more specific estimations of motor timing. Experimental Brain Research, 214(2), 175–183.
Jensen, A. R. (1992). The importance of intraindividual variation in reaction-
cognitive and non-cognitive portions of the total variability time. Personality and Individual Differences, 13(8), 869–881.
in ISIP. Another important question is how ISIP variability re- Jensen, A. R. (1998). The g factor. Westport: Praeger Publishers.
lates to variability in reaction time in elementary cognitive Jensen, A. R. (2006). Clocking the mind: Mental chronometry and individual
differences. Oxford, UK: Elsevier.
tasks that have been widely used in chronometric intelli- Jung, R. E., & Haier, R. J. (2007). The Parieto-Frontal Integration Theory (P-
gence research (Deary, 2000; Jensen, 1992, 2006). Earlier FIT) of intelligence: Converging neuroimaging evidence. The Behavioral
studies have found a substantial commonality between and Brain Sciences, 30(2), 135–154.
Kampa, B. M., Letzkus, J. J., & Stuart, G. J. (2007). Dendritic mechanisms con-
choice reaction time and temporal discrimination and judg- trolling spike-timing-dependent synaptic plasticity. Trends in Neurosci-
ment tasks (Helmbold et al., 2007). A reasonable hypothesis, ences, 30(9), 456–463.
given these findings and the present data, would be that ISIP Lewis, P. A., & Miall, R. C. (2003). Distinct systems for automatic and cogni-
tively controlled time measurement: Evidence from neuroimaging. Cur-
shares some top–down control related variance with reaction rent Opinion in Neurobiology, 13(2), 250–255.
time tasks and general discrimination ability, but that there Madison, G. (2001). Variability in isochronous tapping: Higher-order depen-
are also unique contributions to intelligence that represent dencies as a function of inter tap interval. Journal of Experimental Psy-
chology. Human Perception and Performance, 27, 411–422.
more specific roles of temporal variability for cognition, that
Madison, G., & Delignières, D. (2009). Effects of auditory feedback on the
are distinct from attention and other cognitive control long-range correlation of isochronous serial interval production. Experi-
mechanisms. mental Brain Research, 193(4), 519–527.
106 F. Ullén et al. / Intelligence 40 (2012) 100–106

Madison, G., Forsman, L., Blom, Ö., Karabanov, A., & Ullén, F. (2009). Correla- Singer, W. (1999). Neuronal synchrony: A versatile code of the definition of
tions between general intelligence and components of serial timing var- relations? Neuron, 24, 49–65.
iability. Intelligence, 37(1), 68–75. Troche, S. J., & Rammsayer, T. H. (2009). Temporal and non-temporal sensory
Pessoa, L. (2009). How do emotion and motivation direct executive control? discrimination and their predictions of capacity- and speed-related as-
Trends in Cognitive Sciences, 13(4), 160–166. pects of psychometric intelligence. Personality and Individual Differences,
Rammsayer, T. H., & Brandler, S. (2002). On the relationship between general 47(1), 52–57.
fluid intelligence and psychophysical indicators of temporal resolution Uhlhaas, P. J., & Singer, W. (2006). Neural synchrony in brain disorders: Rele-
in the brain. Journal of Research in Personality, 36, 507–530. vance for cognitive dysfunctions and pathophysiology. Neuron, 52,
Rammsayer, T. H., & Brandler, S. (2007). Performance on temporal informa- 155–168.
tion processing as an index of general intelligence. Intelligence, 35(2), Ullén, F., Forsman, L., Blom, Ö., Karabanov, A., & Madison, G. (2008). Intelli-
123–139. gence and variability in a simple timing task share neural substrates in
Raven, J., Raven, J. C., & Court, J. H. (2000). Manual for Raven's Progressive Ma- the prefrontal white matter. Journal of Neuroscience, 28(16), 4238–4243.
trices and Vocabulary Scales. Section 3: The Standard Progressive Matrices. Unsworth, N., Redick, T. S., Lakey, C. E., & Young, D. L. (2010). Lapses in sus-
San Antonio, TX: Harcourt Assessment. tained attention and their relation to executive control and fluid abili-
Schweizer, K., & Moosbrugger, H. (2004). Attention and working memory as ties: An individual differences investigation. Intelligence, 38(1), 111–122.
predictors of intelligence. Intelligence, 32(4), 329–347.

You might also like