Attention Network Test-Interaction (ANT-I)
Attention Network Test-Interaction (ANT-I): Reliability and Validity in Healthy Older Adults
Study examined the psychometric properties of the Attention Network Test-Interaction (ANT-I) in healthy older adults (N = 173; mean age = 65.4, SD = 6.5).
ANT-I is a variant of the Attention Network Test (ANT), a computer-based tool measuring three attention networks: alerting, orienting, and executive control.
The study aimed to evaluate ANT-I's usefulness as a measurement tool in aging and clinical research.
Reliability was assessed using split-half correlation analyses, and construct validity was evaluated with ANOVAs.
Criterion validity was assessed using hierarchical linear regressions to predict performance on conflict resolution, verbal memory, and retrieval tasks.
Results showed significant reliability for all network scores (alerting, orienting, and executive).
Alerting: r = 0.29
Orienting: r = 0.70
Executive: r = 0.68
ANOVAs confirmed the significance of each network score.
Alerting: 18.3 ms
Orienting: 59.4 ms
Executive: 109.2 ms
Executive network score, along with demographic information, significantly predicted performance on conflict resolution, verbal memory, and retrieval tests.
\beta = -0.165 and \beta = -0.184
The study provides new information regarding the reliability and validity of ANT-I test performance in healthy older adults and insights into its potential utility in clinical research settings.
Introduction to Attention Networks
Attention is not a unitary concept but comprises multiple separate yet interacting components and brain networks.
Posner and Petersen (1990; updated in Peterson and Posner 2012) proposed attention consists of three main networks: alerting, orienting, and executive control.
These networks are defined functionally and anatomically based on task performance and brain activation areas.
Alerting:
Involves achieving and sustaining response readiness to environmental signals.
Associated with brain stem arousal systems, thalamic, frontal, and parietal areas, particularly in the right hemisphere.
Orienting:
Involves selectively allocating attention to stimuli in the environment for high-priority processing.
Associated with the bilateral superior parietal area, temporal parietal junction, and frontal eye fields.
Executive Control:
Involves dealing with interference and resolving conflicts.
Associated with the anterior cingulate and prefrontal areas.
Attention Network Test (ANT) and Attention Network Test-Interaction (ANT-I)
The Attention Network Test (ANT) was developed by Fan et al. (2002) and modified into the ANT-I by Callejas et al. (2005).
The ANT-I measures attention networks and their interactions in a single task.
Each trial includes warning tones and/or visual cues before a central target arrow flanked by irrelevant arrows.
Participants indicate the target arrow's direction by pressing the appropriate key as quickly and accurately as possible.
Subtraction scores between relevant conditions measure the efficiency of the attention networks.
Alerting Measurement:
Measured by comparing trials with or without auditory alerts before the target.
Alerts typically decrease reaction time (RT) and increase error rate (ER).
Orienting Measurement:
Measured by comparing trials with a visuospatial cue followed by the target at the same (cued) or opposite (uncued) location.
Performance is worse in the uncued condition because attention must disengage and reorient.
Executive Attention Measurement:
Measured by comparing trials with congruent flankers (pointing in the same direction as the target) to trials with incongruent flankers (pointing in the opposite direction).
Performance is worse with incongruent flankers due to interference.
Utility of ANT and ANT-I
The ANT and its versions are computer-based tools for measuring attention networks within a 20-minute session.
They have been used in various populations, including non-human primates, children, young adults, older adults, and clinical patients.
The ANT and ANT-I have studied the impact of pathology on attention function in diseases like multiple sclerosis, Alzheimer’s disease, and schizophrenia.
They also study the effects of age and cognitive ability on these networks.
The ANT is not a diagnostic tool but helps understand underlying mechanisms of pathologies by comparing changes in attentional functions.
Psychometric Properties of ANT Versions
Studies have validated the three-component model of attention in young adults, older adults, and patients.
Network subtraction scores are significantly different from zero, even with practice effects.
Network scores are relatively independent but may interact.
Alerting network reliability is lower (r = 0.0–0.8) than the other networks.
Executive network reliability is higher (r = 0.5–0.9).
Lower reliability of the alerting network is attributed to its phasic nature, involving rapid changes in mental and physiological states.
Validity of ANT and ANT-I
Less is known about the validity of the ANT and ANT-I in relation to standardized neuropsychological measures.
Examination of validity is crucial for interpreting ANT results in aging and clinical populations.
The objective of the current project was to examine the reliability, construct validity, and criterion validity of the ANT-I in healthy older adults.
The study examined relationships between ANT-I performance and cognitive processes derived from neuropsychological tests.
Measures were grouped into processes (conflict resolution, verbal memory and retrieval, visual memory, conceptual reasoning, divided attention, and verbal monitoring) based on preliminary factor analyses.
Study Design
The project was part of a larger study on the effects of exercise on cognitive functioning.
Participants were administered the ANT-I and neuropsychological tests at baseline.
The study evaluated the reliability of each network score with split-half correlations and examined construct validity with ANOVA.
Criterion validity was evaluated by submitting process scores to hierarchical regression analyses with ANT-I network scores as predictors after accounting for demographic variables.
The hypothesis was that the executive network would show the greatest reliability and predict performance on conflict resolution, while divided attention would be associated with alerting and orienting scores.
Methods: Participants
Participants were from the Brain in Motion (BIM) study, examining the effects of exercise on cognitive functioning.
The BIM study was an 18-month combined quasi-experimental and prospective cohort study.
The current study examined cognitive and neuropsychological data collected during the pre-intervention phase.
The target population was sedentary men and women aged 55–80 years.
Participants were cognitively intact older adults without active cardiorespiratory disease.
Participants provided informed written consent, and the study protocol was approved by the University of Calgary Conjoint Health Research Ethics Board.
Inclusion criteria:
Engages in less than 30 minutes of moderate exercise 4 days per week or 20 minutes of vigorous exercise 2 days per week.
Body mass index (BMI) of less than 35 kg/m^2
Able to walk independently outside and up and down at least 20 stairs.
No active cardiovascular/cerebrovascular disease or obstructive airway disease.
Non-smoker for at least 12 months.
No major surgery or trauma in the last 6 months.
Free of debilitating neurological disorders.
Written permission from their attending physician.
Montreal Cognitive Assessment (MoCA) score \geq 24
The sample (N = 173) included 91 women (age: mean = 64.7, SD = 6.1 years; education: mean = 15.4, SD = 2.3 years) and 82 men (age: mean = 66.1, SD = 6.9 years; education: mean = 15.7, SD = 3.1 years).
Eight participants were excluded from the original sample (N = 181) because English was their second language.
Methods: Predictor Variables - Attention Network Test-Interaction (ANT-I)
The ANT-I is a Python-based program developed by Callejas et al. (2005).
A 13-inch MacBook Pro controlled stimulus presentation and response collection.
Headphones (Bose Quiet Comfort 15) were used to deliver auditory signals.
Each trial includes warning tones and/or visual cues before a central target arrow flanked by irrelevant arrows.
Participants indicate the target arrow's direction by pressing one of the two arrow keys as quickly and accurately as possible.
Specific subtraction scores between two relevant conditions are used to measure the efficiency of the three different attention networks.
Participants had a practice block (24 trials) followed by six experimental blocks with 48 trials/block.
Auditory signal (tone and no tone), spatial cue condition (cued, uncued, and no cue), and target congruency (congruent and incongruent) were orthogonally crossed.
The 12 possible combinations from each condition were pseudorandomly presented so that there were four trials for each combination in an experimental block.
Participants were instructed to maintain fixation on the fixation cross and to respond as quickly and as accurately as possible.
Feedback following errors was given visually only in the practice blocks.
Reaction time (RT) and error rate (ER) were measured.
Methods: Criterion Variables - Neuropsychological Tests and Measures
Auditory Consonant Trigram (ACT) Test:
Assesses complex attention, including divided auditory attention and working memory.
Participants retain three consonants while counting backwards consecutively from a given number for 3, 9, or 18 seconds.
The number of letters correctly recalled (total correct score) and repeated in error (perseveration score) was calculated.
Symbol Digit Modalities (SDM) Test:
Assesses speed of processing, working memory, and scanning (including visual attention, scanning, and processing speed).
Participants write the numbers associated with specific symbols in the written test and say the associated numbers in the oral test, as rapidly as possible for 90 seconds per subtest.
The number of correct answers was calculated for both the written and the oral tests.
Buschke Selective Reminding (BSR) Test:
Assesses verbal learning and memory.
Participants learn and recall a list of words with a list-learning procedure over multiple trials.
Participants are selectively reminded only of those words that they had not recalled on the immediately preceding trial.
The number of words recalled during the list-learning phase (total score), the number of words identified by multiple choice, and the number of words recalled after a delay (30 minutes) were calculated.
Medical College of Georgia Complex Figures (MCGCF) Test:
Assesses visual memory.
Participants copy a complex drawing, followed by immediate and delayed (30 minutes) recall.
The scores for the number of items correctly copied and then recalled at the immediate and delayed test were calculated.
Delis–Kaplan Executive Function System (D-KEFS subtests):
Several subtests were used for different aspects of executive function assessment.
Card Sorting (CS) Test: Assesses verbal and nonverbal conceptual reasoning, flexibility, and shifting.
Free Sorting: Participants sort cards into two groups with three cards in each group and describe the categorization.
Sort Recognition: Participants identify the categorization rule used by the experimenter.
The number of correct sorts in Free Sorting and description scores for Free Sorting and Sort Recognition was obtained.
Color Word Interference (CWI) Test: Assesses response conflict resolution and inhibitory control.
Color Naming: Participants name the color of patches.
Word Reading: Participants read color words printed in black ink.
Inhibition: Participants ignore the words and name ink colors in which those words are printed.
Inhibition/Switching: Participants switch back and forth between naming ink color and reading the words.
The time participants took to complete each condition was recorded in seconds.
A speed score was calculated by dividing the number of words given by the time it took to complete naming or reading.
Verbal Fluency (VF) Test: Assesses semantic and phonemic fluency and verbal retrieval.
Letter Fluency: Participants produce words that begin with a specific letter of the alphabet as quickly as possible for 1 minute.
Category Fluency: Participants produce words that belong to a designated semantic category (e.g., fruit).
Category Switching: Participants produce words that alternate between two different semantic categories (e.g., fruit and furniture) as quickly as possible for 1 minute.
The number of correct words generated and repeated in error (perseveration score) was calculated for each condition.
Methods: Process Scores
Principal Component Analysis (PCA) was conducted to confirm the factor structure underlying the neuropsychological tests.
PCA grouped measures into independent cognitive processes within overarching domains.
A total of 23 outcomes from the neuropsychological battery were included in the PCA analysis.
The number of appropriate factors was determined using a Monte Carlo Simulator.
Both generated models included a Varimax rotation with Kaiser normalization applied to the entire data set.
The initial eigenvalues showed that the initial factor explained 23.8% of the variance, the second factor 11.7%, the third factor 9.6%, the fourth factor 8.2%, the fifth factor 6.2%, and the sixth factor 5.8%.
The six factors accounted for a total of 65.3% of the variance.
Factor 1: Conflict resolution
Factor 2: Verbal memory and retrieval
Factor 3: Visual memory
Factor 4: Conceptual reasoning
Factor 5: Divided attention
Factor 6: Verbal monitoring
Functions were categorized into three domains: executive function, memory, and complex attention.
Three measures (BSR Test multiple-choice recognition score, VF letter fluency score, and MCGCF Test image copy score) did not load onto any factor and were excluded from further analyses.
Process scores were calculated by averaging the z scores of relevant measures, with polarity changed when necessary for consistent interpretation.
Data Analysis
Preliminary analyses included descriptive statistics for all measures.
Reliability and construct validity analyses replicated Ishigami and Klein (2010, 2011).
Reliability of each network for RT and ER was examined using a variation of a split-half correlation.
Trials were randomly split into two halves 10,000 times, a correlation was computed for each split, and the reliability score was calculated as the mean of the 10,000 correlations.
Construct validity of the network scores was examined by conducting repeated measures analyses of variance (ANOVAs) on mean correct RT and ER with the factors of auditory signal, spatial cue condition, and target congruency.
Strength of the network scores was examined by the presence of main effects.
Isolability was examined by testing for the presence of significant interactions among the networks in the ANOVAs.
Criterion validity was examined using hierarchical linear regressions with age, education, and sex as initial predictors, followed by attention network scores.
Results: Reliability
For RT, all standard network scores were significantly reliable.
Alerting: r = 0.29
Orienting: r = 0.70
Executive: r = 0.68
Reliability of auditory alerting and visual alerting scores was also significant.
Auditory alerting: r = 0.19
Visual alerting: r = 0.36
For ER, none of the reliabilities were significant, possibly due to the small number of errors obtained.
Results: Construct Validity
For RT, the three components were significant, as seen in the main effects of alerting, orienting, and executive control.
Alerting: F(1, 172) = 179.516, p < 0.001
Orienting: F(1, 172) = 431.278, p < 0.001
Executive control: F(1, 172) = 2242.009, p < 0.001
The main effects of auditory alerting and visual alerting were also significant.
Auditory alerting: F(1, 172) = 202.946, p < 0.001
Visual alerting: F(1, 172) = 482.744, p < 0.001
ER results showed similar patterns, with significant main effects obtained for orienting and target congruency.
Orienting: F(1, 172) = 8.280, p < 0.01
Target congruency: F(1, 172) = 32.674, p < 0.001
The interaction between the effects of alerting and orienting in RT was significant (F(1, 172) = 9.590, p < 0.01), with spatial cueing effects greater in the presence of alerting signals.
Interactions between auditory alerting and executive control and between visual alerting and executive control were significant.
Auditory alerting: F(1, 172) = 4.597, p < 0.05
Visual alerting: F(1, 172) = 8.62, p < 0.01
No other interactions were significant either in RT or ER.
Results: Criterion Validity
Two-step hierarchical regression models were used to predict process scores.
Executive Function:
Conflict Resolution:
Demographic variables significantly contributed to the first model (F(3, 167) = 12.47, p < 0.001
Executive network, along with age and sex, was the only significant ANT-I predictor. Greater executive network scores (worse performance) were associated with poorer conflict resolution scores.
Conceptual Reasoning:
Demographic variables significantly contributed to the first model (F(3, 169) = 9.26, p < 0.001
Age and education were significant predictors, but no attention network was significantly associated with the process score.
Verbal Monitoring:
The first model including demographic information was not significant (F(3, 169) = 2.47, p = 0.06
Age was a significant predictor, but no attention network was significantly associated with the process score.
Memory:
Verbal Memory and Retrieval:
Demographic variables significantly contributed to the first model (F(3, 169) = 17.2, p < 0.001
Executive network, along with age, sex, and education, was a significant predictor. Greater executive scores (worse executive performance) were associated with poorer scores in this process.
Visual Memory:
Demographic variables significantly contributed to the first model (F(3, 169 = 6.98, p < 0.001
Age was a significant predictor, but no attention network was significantly associated with the process score.
Complex Attention:
Divided Attention:
Demographic variables significantly contributed to the first model (F(3, 169) = 3.05, p < 0.05
Alerting and orienting networks, in addition to education, were significant predictors. Greater visual alerting scores and smaller orienting scores (i.e., smaller spatial cueing effects) were associated with better scores in this process.
Discussion
The study examined the construct validity, reliability, and criterion validity of performance on the ANT-I in healthy older adults.
The network scores were reliable, valid, and generally independent.
Visual alerting and orienting network scores were related to performance on a test of divided attention.
The executive network score was predictive of performance on tests of conflict resolution and verbal memory and retrieval.
Discussion: Reliability and Construct Validity
The reliability of the alerting network was the lowest, although all network scores showed significant reliability.
The presence of each attention network was confirmed in this large population of healthy older adults.
The alerting effects were associated with facilitation of the orienting effects but a reduction in the executive network.
Discussion: Criterion Validity
Adding scores of all three attention networks at once did not help to predict scores of the neuropsychological test processes, but scores of each network independently had some associations with the neuropsychological test processes.
Alerting:
Larger alerting effects were associated with better performance in the ACT Test, suggesting that larger visual alerting scores indicate better attention.
Orienting:
Smaller orienting scores were associated with better performance in divided attention derived from the ACT Test. Smaller orienting scores could arise when participants are able to efficiently disengage from the cue presented at the location opposite from the target location.
Executive Control:
Larger executive (greater conflict) scores in the ANT-I were associated with poorer conflict resolution.
Larger executive scores in the ANT-I were associated with poorer visual memory and retrieval performance, supporting the involvement of the executive network with the memory components.
The current study provides evidence for the construct validity of the ANT-I by showing relationships between the network scores and process scores derived from the various neuropsychological tests.
Potential Applications and Conclusion
One of the advantages of the ANT-I lies in its ability to provide precise measurements of alerting, orienting, and executive networks in RT (and accuracy).
The results in the current study showed that performance on the attention network measures can be associated with some neuropsychological processes derived from the standardized neuropsychological tests.
The study confirmed that reliability is lowest with the alerting network, the network scores are valid, and the networks are isolable but do not operate independently in all situations, in a healthy older adult population.
Importantly, there were clear links between the executive network and conflict resolution, verbal memory and retrieval, and divided attention.
These results highlight the validity of ANT-I performance with regard to at least some processes derived from standard neuropsychological tests.