Standard Dual N-Back Working Memory Training
The aim of all working memory (WM) cognitive interventions is to expand working memory capacity. The most widely studied brain training exercise targeting WM capacity is the dual n-back task. The N-back task involves viewing a continuous stream of items (e.g., letters) and deciding whether each item matches the stimulus presented n stifmuli back. In dual n-back training, a verbal and a visuo-spatial stream of items is presented simultaneously and item matches have to be detected for both types of information.
This dual task requires updating items in both the visual short term storeand the verbal short term stores of working memory. The most well-known dual n-back task has been developed by Dr. Susanne Jaeggi and her colleagues back in 2008. We call this the ‘Standard’ dual n-back.
Known benefits of dual n-back working memory training
Working memory cognitive interventions result in widespread cognitive benefits (reviews: Morrison & Chein, 2011; Salminen, Strobach & Schubert, 2012):
- Increased performance on untrained measures of short term memory.
- Fluid intelligence (IQ) (see below)
- Multi-tasking – i.e. attentional selection between two sets of information associated with different tasks.
- Detaching attention from irrelevant items and attending to new relevant items.
- Shielding against interfering information.
- Episodic memory.
- Reading comprehension.
- Verbal learning and every day attention in older adults (60+).
- Reduced symptoms of ADHD.
- Improvements for multiple sclerosis – everyday memory, quality of life.
- Improvements for schizophrenia patients – everyday memory, quality of life.
- Improvements for frontal lobe stroke patients.
In general terms, the larger your working memory capacity or ‘mental workspace’, the greater your capacity for higher order cognition and thus academic and professional achievement. An example of this relationship is shown below.
Limitations of Standard Dual N-Back Training
Traditional dual n-back cognitive interventions have a number of critical limitations. My software apps i3 Mindware and HighIQPro have been designed to counter these limitations. [listdot]
- Motivation. N-back training is hard and traditional n-back apps on the market have a high drop out rate because of this. i3 Mindware and HighIQPro have built in motivators that ensure that you complete the program and reap the benefits to your IQ and brain function that would otherwise not be tapped. These include leaderboards, game-like sound effects, block by block customization and IQ increase and top-score guarantees.
- Counter-productive strategies. The original Jaeggi dual n-back and its variants on the market do not counter a number of strategies to maintain or improve n-back levels without associated gains in actual working memory capacity – such as chunking, attention blinking and playing the odds. 2G training (explained below) helps counter these counter-productive strategies.
- IQ gains. With dual n-back training alone, the research is mixed as to whether training generalizes to IQ. A number of recent studies have challenged the Jaeggi study, showing that with certain populations (e.g. college students) n-back training does not result in an improved IQ – such as this one.
2G N-Back Cognitive Interventions
My software apps i3 Mindware (increase IQ) and HighIQPro (improve problem solving) are based on 2G (second generation) n-back training. This consists of the following features.
1. Interference Control
2011-2014 research reveals that interference control is critical to high levels of fluid intelligence and working memory. Interference control is a kind of attention control – the ability to filter out distracting information or suppress irrelevant habits or responses, when faced with cognitive challenges. What is ‘interference’ in the dual n-back? You may have experienced interference in the standard dual n-back when the sequence of stimuli repeats itself before the target is presented. This creates confusion where you have to ‘repeat yourself’ to keep the series of items in memory. The level of interference in 2G training can be systematically increased. This is not possible in standard dual n-back games.
Studies by Burgess, Gray, and fellow grad student Tod Braver (article 1, article 2) provide brain imaging evidence of a large overlap of gF and WM span brain mechanisms when there is need for interference control on a task – but not otherwise.
Why evidence is mixed for IQ gains from dual n-back training?
Some studies such as this one and this one provide evidence for IQ gains from n-back training, while others such as this one and this one do not. Why the inconsistency? An explanation is that none of the studies systematically vary interference in the n-back task. If interference is generally greater in the n-back, transfer to IQ gains will be greater. i3 Mindware and HighIQPro are the only apps that provides users with control over interference level. [hr]
2. Stimulus Speed & Rhythm
Electrical activity in the brain can be recorded using EEG electrodes on the scalp. Different types of brain waves (electrical oscillations) have been linked to sleep, navigation, cognition, attention, and can help diagnose a wide range of disorders including autism, schizophrenia and epilepsy.During standard dual n-back training, while our attention system keeps track of the updating letters and squares, sensory areas of the brain go into an electrical rhythm that matches the rhythm of the n-back letters and squares – for instance, every second. The n-back stimuli drive drive the cortex rhythmically (Lakatos et al., 2009).
The cortical brain rhythm helps process the stimuli, anticipating when the next sounds and squares will appear, and ‘binding’ the audio and visual information together in working memory. How does this relate to real world cognition? Speech is rhythmic, our gestures are rhythmic, visual saccades (the moment to moment eye movements that we scan scenes or text with) are rhythmic. Our attention system works in these rhythms, resulting in periodic increases in excitability in anticipation of attended stimuli – making information processing more selective and efficient. The speed and rhythm-breaks of the stimuli in 2G n-back can be systematically increased. The default is 1 which is a regular 3 second stimulus rhythm. This is the standard Jaeggi n-back setting, and the setting that is used in most commercial and online n-back brain training apps. 2 and 3 settings increase the speed and break the rhythm of the n-back stimuli, requiring more focused attention and demands on working memory.
3. Performance Accuracy
In the original Jaeggi dual n-back algorithm, staying at the same n-back level without dropping down a level requires that you are 75% accurate. Going up an n-back level requires you are 85% accurate. Because there is considerable room for error (such as missing targets) in these accuracy levels, it is possible to use game-specific strategies to improve your n-back score. These strategies do not actually help you expand your brain’s working memory capacity, although they result in increasing n-back levels and the appearance of better working memory performance. The accuracy level in our 2G n-back can be systematically increased. The default is 1 which is the standard Jaeggi accuracy settings. 2 and 3 settings increase the accuracy requirement progressively, with level 3 requiring 100% accuracy to go up an n-back level. Higher accuracy levels prevent game-specific strategies that artificially increase your n-back levels. Using any of these strategies is counter-productive for working memory training. And here’s why. These strategies are actually ways of getting around (i.e. compensating for) limitations of working memory capacity to increase your n-back performance. But getting practice with these strategies does not actually help increase working memory capacity itself. Higher accuracy level settings help prevent the use of domain-specific strategies such as the following:
- Chunking. Sometimes during the standard dual n-back game, a letter or location may be repeated one two or even three times. When this happens it is easier perform on the n-back exercise because with only one ‘place holder’ there is less information to ‘encode’ to do the task. Or at other times, there may be a meaningful string of letters that forms a word or acronym, or the sequence of locations forms a known shape. When items can be grouped together like this, easing the burden on our memory system, this is called ‘chunking’. Chunking can also benefit from practice.
- Attention jumping. As you get more experienced with the standard dual n-back game it is possible to strategically direct your attention in ‘jumps’ to useful strings of letters or square locations in order to maintain or go up an n-back level. Using this strategy, you are not actually updating the items in your working memory continuously, but are ‘counting through’ a particular string of items and then refreshing it from the start again for the next string.
- Playing the odds. Another strategy that can be used in the standard dual n-back is ‘intelligent guessing’ when you are less precise with your location memory, for instance, but can make good guesses that the stimulus was to the left, or to the right for instance. This gives you better ‘odds’ at reach the required level of accuracy to maintain or go up an n-back level.
4. Mindware Strategy Training
i3 Mindware and HighIQPro‘s second generation n-back training combines optional IQ problems and problem-solving training (‘mindware’) with working memory n-back training, to harness a synergy between working memory and learning problem solving strategies. Your working memory capacity is like a computers RAM power. Having more RAM allows you to process more information faster, but it doesn’t help you know what rules and strategies to apply to the information in order to get the right results. It doesn’t help you overcome biases in how you reason, for instance. And it doesn’t help you solve problems unless you know the right strategies to apply to the problem. To learn how to do a mathematical calculation, you need to learn mathematical rules; to learn how to drive a car, you need to learn how to operate the car; to learn a second language, you need to apply grammatical rules. For overall cognitive performance, you need both cognitive capacity (a measure of the efficiency of your brain), as well as the right ‘mindware‘ – knowledge, know-how, rules, procedures and mental strategies.
A larger working memory capacity functions to allow for easier learning and application of mental mindware. Working memory circuitry is located in the lateral prefrontal cortex. Activation in this area allows for the flexible application of a wider range of novel strategies and rules [reference], overcoming automatic responses. These new rules – with practice – are then more effectively encoded in more anterior (to the front of the head) prefrontal areas [reference]. This is shown in the ‘Capacity-Strategy’ infographic below.
Executive Functioning and Episodic Memory
1. Countering the problem of not having multi-modal binding
While the traditional dual n-back task involves the parallel attentional processing in working memory of both visuo-spatial and verbal information, it does not require that the information is integrated in a ‘multimodal’ code. There is no requirement for ‘binding’ of different sources of information – the function of the Episodic Buffer in the model.
But multi-modal, integrated information is arguably more closely related to real world cognitive abilities such as numeracy and problem solving and decision-making. For example, in solving a maths problem, verbally based rules may be combined with graphical visualizations, and in decision-making visualizations may accompany the application of decision-rules. Attending to temporarily integrated multi-modal information is critical for much goal-directed cognition.
I have designed the SuperNBack to train not only independent streams of information in different modalities as in the traditional Jaeggi dual n-back, but also short term conjunctions of visual and audio information too, thus training not only the Phonological Loop and the Visuo-Spatial Scratch Pad but also, critically, the Episodic Buffer.
Training conjunctions as well as disjunctions is expected to greatly increase the effectiveness of n-back training (whether single, dual n-back or, in some versions, triple-n back).
2. Optimizing the learning regime: micro- and macro-level task demands
To optimize the learning regime David, Monty and I have designed the SuperNBack exercise to have a progressive ‘levels’ format, such that once a threshold level of performance are attained on less demanding and complex n-back tasks (e.g. single n-back tasks), the user ascends to the another more complex n-back tasks (e.g. dual n-back, and conjunction n-back tasks).
Thus there are two types of ‘continuous performance’ task demands that evolve together – at a ‘micro’ level of increasing n-back (working memory capacity) performance within a given n-back task, as well as at a ‘macro’ level of increasing the complexity and attentional demands of the n-back task itself – with five different n-back tasks of increasingly complexity.
Both of these challenges acting in concert during training greatly accelerates the rate at which working memory benefits are acquired.
3. Countering the problem of poor incentives and low motivation
One of the problems with the traditional dual n-back and its current variations on the market is that there is a high ‘drop out’ rate before the training program is complete.
To counter this problem, David and Monty have designed a platform for the SuperNBack training task that is highly ‘incentivized’, making use of a virtual private network (VPN) database, for real-time and ‘congratulations’ messaging to the user for performance milestones (whether on a desktop or mobile device), as well as ‘reminder’ texting in case the training is falling behind schedule. Social sharing of training milestones is also facilitated via Facebook and Twitter.
It is expected that with this highly personalized incentive system, the training completion rate will be high, making it an ideal application for institutions with an interest in increasing cognitive performance, as well as for individuals.
Intuitive Decision Making
Objective: to develop an effective training method to improve fast, intuitive decision-making based on probabilities, utilities and signal noise.
Studies indicate that complex, multi-dimensional decisions better made unconsciously/intuitively without the consciously articulating and weighing up pros and cons. When you articulate pros and cons, you set up cognitive biases that reduce the complexity of what is relevant to the decision. Faced with a complex decision, unconsciously the mind is able to process many relevant factors in parallel, apparently resulting in an emerging polarization of pros and cons from which emerges a ‘feeling’ for yes or no, good or bad.
Using signal detection theory I am developing an application which develops skill acquisition of rapid decisions under uncertainty where optimized decision making depends on shifting the ‘internal bias’ of the decision maker, by factoring in:
- the benefits and costs (utilities) of ‘targets’ and ‘foils’
- the probabilities of targets and foils
- the signal noise associated with targets and foils
The OODA decision making application is named after the military strategist Boyd’s Observe-Orient-Decide-Act model: