Do some people have HIGH IQs and POOR working memory?

[Payman Saghafi]

We always hear about the correlation between IQ and working memory.
I'm wondering if there are exceptions to the rule in which a person
with a very high IQ has poor working memory. I have read a little
research about this, but would REALLY appreciate some additional
information.

Pay

[Tistou Blomberg]

That is a really good approach! I would also like to know if that person exists.

[Payman Saghafi]

Is there anybody on this forum that knows anything about this?  There are pages and pages of discussions regarding n-back training in this forum. Surely someone knows of some research that touches on the breadth of working memory prowess in high IQ individuals. 

Somebody?

Pay

[Green]

You might search the literature on TBI to find examples of such people.

[The.Fourth.Deviation.]

I have high numerical/verbal wm but low visuospatial wm. Therefore I scored 100 on the performance scale.but 155 on the verbal scale. My IQ is therefore high, and I have a pronounced wm deficit in a specific domain. However the verbal strength compensates for the deficit.

[Payman Saghafi]

Thank you for the info!  May I ask you a few more questions?  

1)  What was your full scale score on the SB?
2)  Have you taken the WAIS or a RPM type test?
3)  When you took the SAT or ACT was reading comprehension difficult for you? 
4)  Do you consider yourself better at mathematics or English?

Regards,

Pay

On Monday, July 9, 2012 9:41:22 PM UTC-6, krs wrote:

On Thursday, July 5, 2012 2:37:55 PM UTC-5, Payman Saghafi wrote:

I think I'm the guy you're looking for.

On most of the subtests in the SB battery my IQ ranges from the upper-120s to
mid-130s. On the subtests for memory my IQ ranges from the upper-70s to the
mid-90s. People like me are considered learning disabled for this reason.

In comparison, when I took the MAT for grad school my score gave me an IQ-equivalent
in the upper 160s. Getting in to grad school was easy, but getting out was hard.

[Karl Sackett]

On Tue, Jul 10, 2012 at 12:46 PM, Payman Saghafi <payman...@gmail.com> wrote:

Thank you for the info!  May I ask you a few more questions?  

1)  What was your full scale score on the SB?

Sorry, I don't remember. ;)

I think it was around 128.

 

2)  Have you taken the WAIS or a RPM type test?

I took some form of the WAIS when I was in high school. My score was 

high enough for me to qualify for Mensa.

3)  When you took the SAT or ACT was reading comprehension difficult for you?

Reading comprehension has never been a problem with me. Readability, 

however, is a problem. I have both ADD and memory problems. Some books, 

college text books for example, are difficult for me to read, not only 

because of the density of the material, but more because by the time 

I've read to the end of a paragraph I've already forgotten what was 

at the beginning. And this assumes that I've been able to focus my 

attention on the book and not have it fly off in a dozen different 

directions at once.

Expressing what I've read is also a problem. Organizing my thoughts 

in working memory takes time, and if it takes too long I'll forget 

what it was I was trying to express.

  

4)  Do you consider yourself better at mathematics or English?

That's a hard question to answer. What little college math I managed 

to learn has mostly faded away through disuse, but enough is left 

that I can look up the rest in textbooks and on-line to solve what 

problems I run in to. On the other hand I can easily solve math-based 

puzzles and games.

Technical communication isn't a problem. Expressive communication is. 

I can write the manual for running your kitchen, but don't ask me to 

write a review of your restaurant. Also, I've found in spite of my memory 

problems I'm good at linguistics and foreign languages. I taught myself 

Russian using the Pimsleur method, so I can say that spaced repetition

and graduated-interval recall definitely work better for me than 

rote memorization. 

--
Karl Sackett                                      karl.s...@gmail.com

My profiles:
Contact me: karl.sackett

Get a signature like this. CLICK HERE.  

[jotaro]

should i help out too?

i have 128-133 and i am free of the issues mentioned above

i am mainly a visual person.

 

 

 

[paymansaghafi]

Sent from my Verizon Wireless 4GLTE smartphone.

 

---

From : jotaro
Subject : Re: Do some people have HIGH IQs and POOR working memory?
 

should i help out too?

i have 128-133 and i am free of the issues mentioned above

i am mainly a visual person.

 

 

 

 

 

 

--
You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.
To post to this group, send email to brain-t...@googlegroups.com.
To unsubscribe from this group, send email to brain-trainin...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/brain-training?hl=en.

[paymansaghafi]

[Payman Saghafi]

The detail you provided is very generous. Thank you very much!

Pay

[Payman Saghafi]

May I ask how you perform on scholastic standardized tests?  How do you perform on the Raven type tests?

Regards,

Pay

[Payman Saghafi]

Sure, I would like to hear about your story.  Please be honest about anything you tell me.  Thank you!

[Payman Saghafi]

May I ask how you perform on standardized scholastic tests like the ACT or SAT vs. a test like Raven's Progressive Matrices?

On Sunday, July 8, 2012 11:41:15 AM UTC-6, The.Fourth.Deviation. wrote:

[Payman Saghafi]

Nicholas, it is interesting that you are answering this question because I recently posted a related topic:

https://groups.google.com/forum/?fromgroups=#!topic/brain-training/A9tqRQQAmOg

Also, I would love it if you took some of the tests on my web site (including the "underlying test of working memory") that I have created:

http://www.gkacademy.org/    (SEE THE "DIAGNOSTIC TESTS" on the navigation menu on the right side of the page.)

Let me know if you get a chance to try these tests. 

Also, you can have your family members try these tests too as I am curious to see how your working memory compares to your IQ.  The results are confidential.

Regards,

Payman


On Friday, November 16, 2012 3:19:59 PM UTC-6, Nicholas H wrote:

Hello! I come from a family where this is inherited. My sister and I have it and one of her kids have low working memory but high IQ's. I expect both of mine do as well, but it's too early to tell.

 

I am similar to my nephew, I don't know my scores, (though computer tests show between 140-167) but I know his...

Gf or Fluid intelligence = 160's

Gc = 160's - crystalized

Glr= 160's - long term retrieval

Gsm=106 - short term working memory

Gs=110's - maintained focused attention

 

Average was 144.

 

I suspect that some of this is even compensatory, like long term memory is used more in working memory processing and fluid intelligence processing, and that some of the numbers could even be higher and lower. Like another poster, spaced repetition

and graduated-interval recall works better for me as well. Rote is an almost meaningless task for me, but long term spaced repetition leads to mastery for me. I also prefer to understand the full concept and the entire context of something, and I use these understandings in new learnings to build on things and make new learning easier. New concepts that can be connected to old concepts in anyway are extremely easy for me to learn. Entirely new concepts can be extremely hard for me (relatively) to learn and require much longer term exposure. A lot of processing goes on behind the scenes for me.

 

To answer questions about reading and math, I am test well in both. Reading was a harder task for me, it was both slow and I often had to repeat reading paragraphs or even sentences if my mind wandered, only to have forgotten it and have to re read it once I got to the questions, however I am an avid reader and read on a completely different level now, When I read now I go into a kind of a trance where the reading becomes like a movie too me, I have thoughts about the meaning, or the characters, and I have meta thoughts  about the structure and what it is leading up to and how it is conected to everything else I know. My first PSAT was 1040 - 590 on the math and 450 on the reading.  a half later I scored 1340, 650 on the reading and 690 on the math. In the interval I had some kind of break through on reading, I am far more advanced in reading now however (I realize I may have mentioned this before). Originally non-fiction was extremely hard for me, even into college, but it is much more digestible now. I still have difficulties with texts that have a bunch of irrelevancies or too much description (like a fiction book with a bunch of irrelevant details about the scene). For math algebraic abstract math was much easier, Gemetric proof type stuff has always been harder, but once I taught myself to intuit geometry it was much easier, however the proof's are still hard for me.

 

[D Kong]

Very interesting. I am somewhat similar...

I scored a 12 on the working memory test, Payman. Although I didn't manage my time very well and am not in great performance, but it does reflect that I struggle with working memory. 

It took me a couple of days to even really grasp dual n-back.

--

You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.
To post to this group, send email to brain-t...@googlegroups.com.
To unsubscribe from this group, send email to brain-trainin...@googlegroups.com.

Visit this group at http://groups.google.com/group/brain-training?hl=en.
 
 

[Pontus Granström]

Several investigators have claimed over the past decade that working memory (WM) and general intelligence (g) are identical, or nearly identical, constructs, from an individual-differences perspective. Although memory measures are commonly included in intelligence tests, and memory abilities are included in theories of intelligence, the identity between WM and intelligence has not been evaluated comprehensively. The authors conducted a meta-analysis of 86 samples that relate WM to intelligence. The average correlation between true-score estimates of WM and g is substantially less than unity ( =.479). The authors also focus on the distinction between short-term memory and WM with respect to intelligence with a supplemental meta-analysis. The authors discuss how consideration of psychometric and theoretical perspectives better informs the discussion of WM–intelligence relations.

[Pontus Granström]

Abstract

Investigates the relationship between three factors of working memory (storage and processing, relational integration, and supervision) and four factors of intelligence (reasoning, speed, memory, and creativity) using structural equation models. Relational integration predicted reasoning ability at least as well as the storage-and-processing construct. Supervision, measured as specific switch costs, was not related to intelligence, but general switch costs were moderately correlated to the reasoning factor. The results question the view of working memory as a device for storage and processing, and the executive-attention account of working memory. They are better explained by theories describing working memory as a system for building relational representations through temporary bindings between component representations.

Table 3. Correlations between latent factors of intelligence and working memory, and regression weights for working memory factors predicting intelligence factors

	Reasoning	Speed	Memory	Creativity
Correlations
Storage and processing	.81	.40	.66	.28
Relational integration (full)	.94	.64	.58	.47
RI-no memory	.88	.63	.59	.48
RI-memory	.92	.51	.52	.40
Supervision (SSwi)	(.21)	(.13)	(.18)	(− .17)
Supervision (GSwi)	.34	(.15)	.34	(.07)
Regression weights WMC → Intelligence
Storage and processing	(.28)	(− .18)	(.59)	(− .17)
Relational integration (full)	.71	.80	(.06)	.64
Supervision (SSwi)	(− .10)	(− .05)	(− .03)	− .33
Supervision (GSwi)	(.08)	(− .08)	(.25)	(− .11)

[Payman Saghafi]

Yes, I have seen that meta analysis before, King of Stars.  It is a good one.  :)

[Pontus Granström]

The correlational analyses shed further light on the nature of the relationship between working memory and general ability. The relationship between the two factors was stronger in the low ability group compared to the high ability students. This finding is not inconsistent with the view that IQ or g functions like a central processor (Anderson, 1992). In low ability groups, this central processor has to work harder on all cognitive tasks, which may explain why there was a stronger relationship between working memory and IQ tasks. In contrast, with high ability students, the central processor does not have to work as hard and so working memory is not as constrained by performance in IQ tests (see Reynolds & Keith, 2007).

[Pontus Granström]

here are several implications for the current findings. First, it suggests that alternative assessments of cognitive skills, like working memory, might yield accurate estimations of ability. Some suggest that the reliance on IQ scores to identify high ability children can be problematic due to discrepancies in performance between verbal and nonverbal tests (Sweetland, Reina, & Tatti, 2006). Furthermore, IQ tests are sensitive to socioeconomic factors such as maternal education level (Groth, 1975), caregivers' attitude towards education (Reynolds, Willson, & Ramsey, 1999), and cultural differences (Brody & Flor, 1998), which may result in an under-representation of children from lower socioeconomic backgrounds in gifted programs (Borland, 2009). In contrast, working memory appears to be relatively impervious to such factors like maternal education (Alloway et al., 2005) and income levels (Engel, Santos, & Gathercole, 2008). Working memory skills also seem to be a robust predictor of academic attainment, even when IQ is statistically accounted for ( [Alloway, 2009] and [Alloway and Alloway, 2010]). Thus, standardized working memory assessments can provide a suitable alternative for classification of ability levels.

Quite frankly I am quite tired of the "Anti working memory movement" and all their logistical and almost religious explanations.

[Pontus Granström]

Fluid reasoning shares a large part of its variance with working memory capacity (WMC). The literature on working memory (WM) suggests that the capacity of the focus of attention responsible for simultaneous maintenance and integration of information within WM, as well as the effectiveness of executive control exerted over WM, determines individual variation in both WMC and reasoning. In 6 experiments, we used a modified n-back task to test the amount of variance in reasoning that is accounted for by each of these 2 theoretical constructs. The capacity of the focus accounted for up to 62% of variance in fluid reasoning, while the recognition of stimuli encoded outside of the focus was not related to reasoning ability. Executive control, measured as the ability to reject distractors identical to targets but presented in improper contexts, accounted for up to 13% of reasoning variance. Multiple analyses indicated that capacity and control predicted non-overlapping amounts of variance in reasoning.

KEYWORDS:

working memory, attentional capacity, executive control, fluid reasoning, n-back task


Fluid reasoning tests are believed to give a good measure of general fluid intelligence (Gf),1 which reflects the inter-individually varied but intra-individually constant unacculturated ability for abstract reasoning on novel material. In terms of the material's novelty, Gf differs from crystallized intelligence (Gc), which involves the effective application of previously acquired knowledge (Cattell, 1971). Gf is closely related to the concept of general intellectual ability (Spearman's g; Gustafsson, 1984), and both Gf and g factors are significant predictors of professional and personal success in human life (Gottfredson, 1997).

Working memory (WM) is a basic cognitive mechanism responsible for the active maintenance of information for its ongoing processing. The construct of WM, introduced to psychology by Baddeley and Hitch (1974), refers to a kind of cognitive “engine” involved in various mental processes. It is usually believed to be responsible for holding and manipulating temporary solutions, structures, subgoals, or subproducts of thinking, before the final result is reached. Its most important feature is its severely limited capacity (Cowan, 2001; Daneman & Carpenter, 1980).

Over the last 20 years numerous studies (e.g., Colom, Abad, Rebollo, & Shih, 2005; Conway, Cowan, Bunting, Therriault, & Minkoff, 2002; Engle, Tuholski, Laughlin, & Conway, 1999; Kane et al., 2004; Kyllonen & Cristal, 1990; Martínez et al., 2011; Süß, Oberauer, Wittmann, Wilhelm, & Schulze, 2002) have proved that fluid reasoning is strongly correlated with working memory capacity (WMC). Analyses of the results of multiple psychometric studies estimate that WMC and Gf share from 50% (Kane, Hambrick, & Conway, 2005) to more than 70% (Oberauer, Schultze, Wilhelm, & Süß, 2005) of variance, thus making WMC the strongest single predictor of fluid reasoning. Consequently, most WM researchers (e.g., Conway, Kane, & Engle, 2003; Oberauer, Süß, Wilhelm, & Sander, 2007) purport that the cognitive processes underlying and constraining WMC also determine the efficiency of fluid reasoning captured by Gf tests (but see Ackerman, Beier, & Boyle, 2005, for a contrasting opinion). So, an important goal for cognitive science is to understand the cognitive and biological mechanisms of WM and explain why people who excel in working memory tasks also easily solve demanding intellectual tests, while others fail in both.

Two main strands of theoretical proposals considering the WMC–Gf link are most influential. Proponents of the first (“a capacity approach”) suggest that both WMC and reasoning depend on the amount of information that can be maintained and integrated within WM by some kind of attentional capacity. Researchers taking the other stance (“a control approach”) believe that both WMC and Gf factors are determined by the efficiency of control over the information kept in WM. Both the capacity and the control approaches enjoy considerable empirical and theoretical support.

However, most of the above cited studies on the Gf–WM link have relied on batteries of relatively complex tasks (e.g., so-called complex span tasks that combine memory storage with some extra processing; Daneman & Carpenter, 1980). Therefore, it is difficult to interpret unequivocally the observed links as reflecting capacity, control, or both. The present research attempts to prove that comparably high correlations between Gf and WM measures may also be observed with a much simpler task, and the various scores on that task may be better interpretable in terms of either capacity or control. Moreover, the crucial methodological aspect of this study is an experimental manipulation regarding the task, which makes WM–Gf correlations appear and disappear. Because many cognitive tasks correlate with Gf due to the phenomenon of positive manifold (Van Der Maas et al., 2006) or coincidence of task requirements (e.g., task complexity, instruction, need for vigilance, novelty of situation), the identification of cognitive mechanisms generating such correlations is difficult. If by application of a specific experimental manipulation we are able to control the strength of correlation between WM task performance and Gf, we can infer that this very manipulated variable reflects the specific aspects of a WM mechanism that are related to Gf (also see Hambrick, Kane, & Engle, 2005).

The general aim of the present article is to use a simple WM task and the manipulations influencing the strength of the WMC–Gf correlation in order to simultaneously examine the involvement of attentional capacity and control processes in operation of WM and to test if both the capacity and control approaches to the explanation of fluid reasoning can be supported.
Cognitive Basis of Fluid Reasoning



The idea that some attentional capacity determines intelligence is obviously not a new one. Its roots may be found in Spearman's (1927) “mental energy” metaphor of the g factor, which described it as a kind of resource that is supposed to saturate all mental activities to some extent. The most thorough examination of attentional limitations within WM was carried out by Cowan (1995, 2001). On the basis of evidence from behavioral, psychometrical, developmental, neurobiological, and formal modeling studies, he argued that working memory capacity is determined by the limit of WM's focus of attention, also referred to as primary memory capacity or attentional scope. The focus is structurally distinct from the other mechanism contributing to WM operation, namely, the activated part of long-term memory (LTM), also called secondary memory. Capacity of the focus is defined as the number of attentionally activated chunks of information easily available for direct use by ongoing cognitive processes. The proper estimation of the number of chunks that attention can span during a given experimental situation depends on the reduction of chunk associations and the elimination of mnemonic strategies, which exploit activated LTM and can artificially inflate the amount of available information. According to Cowan (2001), the average capacity of the focus in healthy adults is four chunks, although it varies among people from two to six chunks, and an individual's span is closely related to Gf level. Two studies by Cowan (Cowan et al., 2005; Cowan, Morey, Chen, & Bunting, 2007) confirmed the close relationship between capacity estimates and intelligence and maturation.

Duncan et al. (2008) also proposed that the Gf level results from some limited capacity, but they explain the nature of this capacity in a somehow broader sense. According to them, WMC does not reveal the capacity limit of a particular storage buffer, but rather it reflects the total capacity of the cognitive system to code multiple demands and components of a novel task. When such a task is complex, for instance, while working on difficult fluid reasoning test items, its different components compete for attention and the most vulnerable ones may be lost, leading eventually to some aspects or goals regarding the task being neglected (Duncan, Emslie, Williams, Johnson, & Freer, 1996).

Halford and his collaborators (Halford, Cowan, & Andrews, 2007; Halford, Wilson, & Phillips, 1998) agreed that adult humans can process about four independent entities in parallel but disagree that this limitation is the sole storage constraint. The authors proposed that the nature of WM capacity limitation consists of the maximum number of processing dimensions (objects, variables, etc.) that can be simultaneously interrelated. Such a number defines the relational complexity imposed by a task. The higher this relational complexity, the more complex representations have to be formed in WM. In case of difficult relational problems used in fluid reasoning tests, the size of representation can easily exceed the available capacity of the majority of people. In accordance with Halford's and Hummel and Holyoak's ideas, Oberauer et al. (2007) proposed that reasoning ability is determined by the number of elements that can be simultaneously bound to a task-relevant structure, such as positions in a mental array during some spatial task or time tags during a serial recall task. These elements can be directly accessible within an active component of WM. Variance in the capacity for such bindings is reflected by WM measures. Additionally, such a capacity determines success in reasoning tasks, because they require the binding of numerous elements into an appropriate mental structure, as well as quick and accurate access to this structure when necessary (Oberauer et al., 2007; Oberauer, Süß, Wilhelm, & Wittman, 2008).

Executive control (or cognitive control) processes are believed to be responsible for the organization and coordination of other mental states and processes in accordance with the internal goals of an individual (Monsell & Driver, 2000). Numerous studies have shown significant correlations between fluid reasoning and the indices of cognitive control obtained from various tasks, for example involving updating (Friedman et al., 2006; Gray, Chabris, & Braver, 2003; Salthouse, 2005); the inhibition of salient distractors, unwanted thoughts, or prepotent responses (Brewin & Beaton, 2002; Dempster & Corkill, 1999; Gray et al., 2003; N cka, 1999); and dual-task coordination (Ben-Shakhar & Sheffer, 2001; Chuderski & N cka, 2010). Although there is no consensus yet concerning how many distinct types of executive processes really exist (see Collette & Van der Linden, 2002; Engle & Kane, 2004; Friedman & Miyake, 2004; Miyake, Friedman, Emerson, Witzki, & Howerter, 2000; Nigg, 2000), nor what the neurobiological mechanisms of control look like (Braver, Gray, & Burgess, 2007; Duncan et al., 2000), the proponents of the control approach agree that the quality of human executive control plays the crucial role in higher cognitive processes.

The most influential theory about the link between executive control, WMC, and fluid intelligence is that of Engle, Kane, Conway, and their collaborators (Engle & Kane, 2004; Kane & Engle, 2002). According to these authors, an executive process determining an individual's effectiveness in both WM and reasoning tasks relies on the proper use of domain-general attentional control, consisting of focusing attention on crucial task-relevant information, rather than spanning attention on all available information. Such a process allows for goal-directed behavior, especially under interference, distraction, or conflict. The proposed link between attention control and WMC has been verified by numerous executive tests, for instance proving that high-WMC individuals, compared to low-WMC individuals, were faster and more accurate on antisaccades (Kane, Bleckley, Conway, & Engle, 2001; Unsworth, Schrock, & Engle, 2004), produced smaller error rates in incongruent trials using a high-congruent version of the Stroop test (Kane & Engle, 2003) and the flanker task (Heitz & Engle, 2007), and more effectively suppressed distractors in a dichotic listening task (Conway, Cowan, & Bunting, 2001).

Most studies that related attentional control to fluid intelligence exploited latent variable modeling. Their results indicated that path coefficients between latent variables reflecting Gf and complex span tasks were equal to r = .49 (Engle et al., 1999), r = .52 (Kane et al., 2004), and could even reach r = .60 (Conway et al., 2002). However, these data should be interpreted with caution, as calculation of the control latent variable required additional assumptions. For example, Kane et al. (2004) assumed that this variable represented the variance common for both simple and complex span tasks, while in the two other studies cited, this variable reflected the variance unique to complex span tasks (i.e., after storage variance had been partialled out).

Additional support for the control approach comes from two studies showing that the Gf–WMC correlation does not depend on the memory load imposed by WM tasks (Salthouse & Pink, 2008; Unsworth & Engle, 2005). In the latter study, correlation coefficients between WM task performance and Gf were similar for two-item and seven-item memory set size conditions, suggesting that it was not the storage capacity factor that determined the value of these correlation coefficients.

Also some neuroimaging studies (Burgess, Braver, & Gray, 2006; Burgess, Gray, Conway, & Braver, 2011; Gray et al., 2003) suggested a strong correlation between reasoning scores and the neuronal activity of the postulated control mechanisms, like the prefrontal or anterior cingulate cortices, during performance in WM tasks. Also electroencephalography (EEG) data have shown that high-WMC brains are better than low-WMC ones in filtering irrelevant items out of WM (Vogel, McCollough, & Machizawa, 2005) and the latter more slowly recover from attentional capture (Fukuda & Vogel, 2011).

A promising research examines the joint contribution of both capacity and control limitations to fluid reasoning. For example, Cowan et al. (2007) proposed that attentional mechanism is flexible, and depending on task requirements it can zoom out to span as many items as possible (e.g., in a recall task), or it can zoom in and focus on a single goal in order to protect it against distraction (e.g., in a task requiring resolving interference). Cowan, Fristoe, Elliott, Brunner, and Saults (2006) showed that two measures, one loading the scope of attention and the other capturing attentional control, shared 12% of variance in intelligence, although scope and control contributed separately to an additional 15% and 10%, respectively. Furthermore, two studies demonstrated the interactive effects of memory load and cross-mapping interference on reasoning (e.g., Cho, Holyoak, & Cannon, 2007; Chuderska, 2010). All these results suggest that the partially overlapping mechanisms responsible for both control and capacity may be the cornerstone of fluid reasoning (see also Schweizer, Moosbrugger, & Goldhammer, 2005).
Measurement of Working Memory



The important issue considers how to properly measure the effectiveness of WM. Traditional short-term memory (STM) tasks have long been believed to mostly involve automatic storage mechanisms (e.g., the phonological loop) and yield low correlations with reasoning (see Unsworth & Engle, 2007). So, WMC was often measured with various versions of complex span tasks (Daneman & Carpenter, 1980). Although these tasks have useful psychometric properties and predict Gf (Conway et al., 2005), as noted in the introduction, they are too complex to be useful tools for understanding how WM mechanisms underlie reasoning. Fortunately, recent studies have shown that also some simple WM tasks can be proper measures of WM, substantially correlating with reasoning. For example, Oberauer, Süß, Schulze, Wilhelm, and Wittmann (2000) demonstrated that simple span tasks, which involve memorizing spatial relations, revealed both high loadings on WM factor and WM–Gf correlations comparable to those obtained with complex span tasks. Kane et al. (2004) showed that tasks requiring only the storage of spatial information correlated stronger with Gf measures than did verbal complex spans. Even simple verbal spans correlated substantially with reasoning, provided that memory load was high enough (Unsworth & Engle, 2006) or rehearsal and chunking were successfully blocked (Cowan et al., 2005).

Two relatively simple WM tasks, which successfully predicted intellectual abilities (Cowan et al., 2005, 2007; Friedman et al., 2006; Gray et al., 2003; Hockey & Geffen, 2004; Kane, Conway, Miura, & Colflesh, 2007; Roberts & Gibson, 2002; Salthouse, 2005; Schmiedek, Hildebrandt, Lövdén, Wilhelm, & Lindenberger, 2009), are an n-back task (Kirchner, 1958; Mackworth, 1959) and a running memory task (Pollack, Johnson, & Knaff, 1959). The former requires matching the current item in a continuous stream of stimuli with an item that occurred n items ago, while the latter requires recall of the most recent items from a serially presented list of unpredictable length. In the present study, we decided to design a task that combines advantages of both these paradigms.

Although the validity of the n-back task as a WM measure has been questioned, as it was unrelated to complex span tasks and both tasks accounted for unique amounts of Gf variance (Kane et al., 2007; Roberts & Gibson, 2002), that fact might simply result from superficial differences in both tasks' requirements (i.e., recognition vs. recall) or the use of single tasks. When either the recall version of the n-back task was used (Shelton, Metzger, & Elliott, 2007) or complex spans and n-back scores from more than one task were aggregated (Shamosh et al., 2008), they shared a quarter of variance. Moreover, in two recent studies, latent variables representing memory updating, which included n-back task scores, were related to variables reflecting either complex-span (Schmiedek et al., 2009) or simple-span (Chuderski, Taraday, N cka, & Smoleń, 2012) tasks, and the observed correlation was close to unity. In our opinion, the most “non-WM” feature of the standard n-back task is the requirement of continuous responding after each stimulus presentation. This makes the n-back procedure more a decisional than a WM task. We believe that even a standard recognition version of the n-back task can be a suitable measure of WM, provided that it is properly modified in such a way that participants do not have to engage in continuous decision making, so they can focus on processes specific to WM, like memory encoding, updating, and searching.

So, in our study we used a modified version of the n-back task, which involved relatively rare responding as in the running memory task, but—unlike the latter task—which relied on a recognition paradigm (i.e., there was no interference from recall procedure). Such a task may be administered in two ways (McElree, 2001). Within a so-called inclusion condition, a participant is asked to retain in memory n the most recent stimuli out of a stream of stimuli and to indicate if a current stimulus (a target) in the stream matches any of these n stimuli. By increasing n values of item repetitions, the load on attentional capacity can be intensified until its limit is exceeded. Cowan (2001, p. 89) noticed that such a task could appropriately be taken as a measure of capacity limits, and in the present article we aimed to test that expectation. Within a so-called exclusion condition, a participant is asked to retain n of the most recent stimuli but to indicate only when a target matches the n-back stimulus. Increasing n also increases the load on attentional capacity, but additionally such a task enables the imposition of a significant load on control mechanisms. This goal is reached by presenting distractor items (so-called lures), which are identical to items placed at positions different than n. Lures activate the tendency for accepting them (i.e., committing false alarms), which must be suppressed. So, the false alarm rate is commonly believed to be an index of the (in)efficiency of control (e.g., Burgess et al., 2011; Gray et al., 2003). Thus, using the exclusion version, the contribution of both target and lure conditions to the prediction of reasoning scores can be compared. This was also examined in the present article.
Goals and Rationale of the Study



Assuming that WM is a simpler cognitive mechanism underlying a more complex process of reasoning, the specific goals of the present article may be defined as the search for both capacity limits and executive control constraints of WM that contribute to the effectiveness of fluid reasoning. In order to achieve this goal, two questions should be precisely answered.

The first question concerns the kind of WM capacity related to reasoning: whether it is the total capacity of WM or the capacity of a particular part of WM (the focus of attention). We assumed that participants would often maintain a small number of recent items in their focus of attention, while less recent items would be usually moved outside the focus (i.e., to the activated part of LTM). Taking into consideration the results of Cowan (2001) and Usher, Cohen, Haarmann, and Horn (2001), who estimated the attentional capacity to around three or four items, we expected that participants would often keep within their foci of attention items up to the 2- or 3-back position (i.e., the currently presented stimulus, the 1- and 2-back items, and sometimes the 3-back item), but they would rarely be able to hold the 4- or 5-back items (ns larger than five were not examined as they usually lead to floor effects) within their attention, because it would require the capacity of five or six items, respectively. Thus, if it is the higher capacity of the (already very limited) focus of attention that makes people more intelligent, reasoning scores and target hit rates will be more strongly (positively) correlated at small n positions (especially at the 2- and 3-back) than at large n positions (i.e., 4- and 5-back). Alternatively, if fluid ability depends on the total working memory capacity expressed as the maximum number of items that can be retrieved from more than one memory area (see Duncan et al., 2008; Unsworth & Engle, 2006, 2007), then one may expect the strongest correlations in cases of large values of n. Another possible result is that no effect of n factor on the Gf–WM correlation would be observed, suggesting that more intelligent participants outperform less intelligent ones generally, for reasons that the present study would be unable to identify decisively but that could be in concord with the control approach (Unsworth & Engle, 2006).

The empirical evidence regarding WM–Gf correlations as the function of memory load seems to be ambiguous. Some data show that significant correlations between WM task performance and fluid reasoning appear only when memory load is high. Unsworth and Engle (2006) analyzed correlations between Raven's score and simple spans at different memory loads and found that correlations increased with rising memory sets, from r = .12, in case of a two-item memory load, to r = .45, for loads of seven items. In cases of tasks more similar to ours, Hockey and Geffen (2004) used the visuospatial n-back task with 0- (i.e., matching predefined target), 1-, 2-, and 3-back exclusion conditions. The authors observed increasing WM-reasoning correlations (rs from non-significant to .27) as a function of n. However, their result could be influenced by the ceiling effect (94% correct) for 0-back and 1-back items. Within the letter n-back task, Kane et al. (2007) observed significant correlations with Raven's test only for the 3-back exclusion condition (as did Gray et al., 2003) but not for the 2-back one. Two studies reported opposite findings. Salthouse (2005) found significant correlations with Gf for 0-, 1-, and 2-back conditions of a digit n-back task (he did not test higher values of n). When the running memory task was applied, the recall from the 2-back position correlated significantly with Gf, while the result from the 3-back position did not. However, Salthouse's data may have been caused by the fact that his research involved elderly participants (up to 95 years old). Also Friedman et al. (2006) reported a significant correlation (r = .28) with Gf for the spatial 2-back task.

In our pilot study (Chuderski & Chuderska, 2009), with the n-back task and an analogical reasoning test, we found the largest difference in WM scores between high-Gf and low-Gf participants in case of the values of n equaling one, two, and three, but not in case of the value of n equal to four. Therefore, we expect to obtain a similar result in the present study. This outcome would support the hypothesis that the effectiveness of reasoning is related to some very capacity-limited part of WM (most probably the focus of attention), but not to the total capacity of WM.

The second question regards whether executive control over WM, which is known to be significantly related to reasoning (e.g., Burgess et al., 2011; Gray et al., 2003), contributes to Gf jointly with WM capacity, with them both predicting overlapping amounts of variance in reasoning, or whether they contribute independently, predicting distinct amounts of Gf variance. If control and capacity contribute to shared variance, and accepting the aforementioned assumption that differences in Gf-related capacity are captured only by small n conditions, then one should expect stronger (negative) correlation between Gf and the false alarm rate in cases of these n conditions than in cases of larger values of n. This would mean that larger capacity somehow facilitates lure rejection, simply because information that a lure is at a position different than n would be more precisely represented in the focus of attention than outside the focus. On the contrary, if control and capacity independently contribute to reasoning then one should expect no effect of n on the correlation between false alarm rate and Gf. By systematically manipulating the n and target/lure conditions of our n-back task, we aimed to test these two opposing hypotheses on (joint vs. independent) contribution to fluid reasoning of WM capacity and control over WM.

Furthermore, the complex pattern of correlations between Gf and WM measures—if found—should reveal not only the cognitive basis of fluid reasoning, but it may tell us something about the structure of WM itself. If some aspects of WM correlate with reasoning but some others do not, then there are good arguments for functional dissociation of these aspects.

We present six experiments. The first three experiments use the inclusion version of the n-back task and investigate the effect of n on the strength of correlation between the target hit rates and reasoning scores. These experiments aimed to provide an answer to the first question. The next three experiments exploit the exclusion version in order to, first, generalize that answer onto this version and, second, to test the pattern of correlations among reasoning scores and target hits and false alarms for varied n positions, and so to provide the answer to the second question.
Experiment 1



The first experiment served to validate our modified n-back task by testing if there would be any effect of increased n on the task performance. Regarding fluid reasoning, we attempted to observe how the value of n would influence the correlation between reasoning and that performance.

We adopted the modified inclusion n-back task from the study by McElree (2001). In his study, McElree used sequences of six- to 15-letter stimuli; a fast presentation rate (900 ms per stimulus); 1-, 2-, and 3-back conditions; and a two-choice response set. The target was always the last stimulus in a sequence of an unpredictable length. We made two major changes to the task to make it more suitable for our purposes. First, in order to capture the process of information maintenance unaffected by mnemonic strategies, we used two-digit numbers as stimuli, because their rehearsal takes longer and their chunking is more difficult. Second, we applied the go/no-go methodology: Participants were instructed to respond only when they detected item repetition and to refrain from responding when the current stimulus did not match any recent item. As a result, participants did not have to respond in every trial and so they could focus on processes specific to WM.
Method
Participants

A total of 101 students recruited from several colleges in Lodz, Poland, participated in the study (44 females; M age = 21.44 years, SD = 2.11; age ranged from 18 to 34 years). For a 2-hr session each person received the equivalent of about €5 in Polish zloty plus a CD-ROM encyclopedia. In this and all subsequent experiments, all participants reported normal or corrected-to-normal vision, they were instructed to take the most comfortable sitting position during the experiment, they completed WM and Gf tests in groups of two to five people, and during the computerized task they were equipped with headphones.
Materials and procedure

Stimuli were 72 out of 90 possible two-digit numbers, excluding nine numbers containing zero as well as nine palindromes. Each stimulus was 1.5 × 1.0 cm in size and was presented in the center of a PC computer screen, in black on a green background. In each trial, a sequence of 4 to 12 stimuli (randomly) was serially displayed for 1,000 ms each. The last stimulus in a sequence (a target) was always identical to an item presented randomly two, three, or four (i.e., the values of n) items ago. There was only one such repetition in a given trial. After the last item in each sequence had been displayed, a black square 1.5 × 1.0 cm in size was presented for 1,000 ms, in order to inform participants that the recent stimulus had been a target and that the next trial was about to start. Each session consisted of nine trials, three trials at random per each n value, starting with a fixation point presented for 1,000 ms. All nine targets differed from one another, and none of the remaining 60 items could match any other stimulus in the whole session. In each session, a presentation of one trial per each possible sequence length (i.e., in the 4–12 range) was intended. However, due to an error in the software controlling the experiment, there was one additional six-stimulus trial, but there was no nine-stimulus trial. The sequence of trial lengths was also random. There were 10 sessions in the test, all followed by short breaks. Two training sessions preceded the experiment.

The participants were instructed to track the four most recent numbers and to press the space bar only when they detected that the current item was identical to an item presented two, three, or four items ago. They were required to refrain from responding in any other case. Only responses that occurred more than 200 ms after a target appeared on the screen and within 200 ms of a target disappearing from the screen were accepted as correct hits. Participants' reactions out of this time frame, together with reactions to non-repeated items, were treated as errors. Errors were signaled with a beep sound.

The inclusion condition does not contain any lures, and the errors cannot be assigned to any particular value of n, so only a total error rate is reported. As these errors constituted a kind of false alarm error, henceforth we will call these errors false alarms to non-repeated stimuli. This will help us to distinguish them from false alarm errors committed in the cases of lures (introduced in Experiments 4–6), which will shortly be referred to as false alarms. False alarms to non-repeated stimuli reflected a bias toward responding in cases when the participants were uncertain whether a non-repeated stimulus had or had not been repeated. The rates of these false alarms were subtracted from hit rates in order to get an unbiased estimate of individual accuracy in responding to targets. Such a correction accounted for the fact that a participant's probability of guessing to respond when she or he was uncertain whether the current trial consisted of a repetition or not did, respectively, increase her or his hit rate in comparison to the “real” hit rate she or he would score if she or he responded only when being certain.

Before the computerized WM task, each participant was given the standard paper-and-pencil test of fluid reasoning: Raven's Progressive Matrices Advanced Version (Raven, Court, & Raven, 1983). The total number of correctly solved items (out of all 36 items) within 40 min was taken as an estimate of an individual's reasoning ability (the reasoning score).
Results

Reliabilities of the consecutive hit rates of the n-back task, estimated with Cronbach's alpha, were α2 = .80, α3 = .69, and α4 = .70 (the subscripts indicate respective values of n). Raven's reliability (calculated from a larger sample of 1,129 participants) was α = .87. As this article focuses on correlational analyses, relevant means and standard deviations for consecutive values of n for this and all subsequent experiments are presented in the Appendix. The rate of false alarms to non-repeated stimuli was relatively low (M = 0.08, SD = 0.05), and for further analyses it was used in order to correct the individual hit rates. Repeated-measures analysis of variance (ANOVA) revealed the main effect of n factor, F(2, 200) = 128.53, η2 = .56, p < .001, indicating that accuracy significantly decreased with increasing n. All three hit rates were strongly correlated (rs > .61, ps < .001).

Gf level correlated significantly with the 2-back (r = .33, p = .001) and 3-back (r = .25, p = .013) hit rates, while the correlation of the reasoning score with the 4-back hit rate (r = .10, p = .317) was not significant. A respective r value for the 2-back condition was significantly higher than the one for the 4-back condition, t(99) = 2.37, p = .001.
Discussion

Experiment 1 yielded two important results. First, increasing the value of n largely decreased performance, replicating the results commonly observed within the n-back paradigm. This effect suggests that 3- and 4-back trials imposed a high load on WM. Second, the significant difference in correlation values between Gf and the 2- and 4-back hit rates is consistent with the hypothesis predicting that the Gf–WM correlation is highest at small values of n. Following our assumptions from the introduction, we interpret this result as indicating that in the majority of trials most participants were generally unable to hold the 4-back targets within their foci of attention, and the more capacious attention of some of them probably did not help them to recognize these targets. On the contrary, the 2-back targets, and probably the 3-back ones, were more often kept within the focus, and the supposedly more capacious attention of more intelligent participants contributed to their better recognition of these targets in comparison to the less intelligent ones. The reliabilities of respective n-back conditions were satisfactory and—most important—comparable, so they could not account for the observed differences in correlation strengths.

However, the main problem with this first study was that the stimulus presentation rate was probably too fast for the applied compound stimuli, resulting in a relatively low hit rate in comparison to the accuracy observed in studies on the n-back task administered with longer presentation periods (e.g., Hockey & Geffen, 2004; Kane et al., 2007). Elsewhere, one of us has shown (Chuderski, Stettner, & Orzechowski, 2008) that the differences in the use of attention that appear within the Sternberg task depend on time pressure. So, in order to reject the possibility that substantial time pressure narrowed participants' actual capacity, more time for responding was allowed in the next study. We also wanted to test if the 1-back hit rates would be related to Gf.
Experiment 2


Method
Participants

A total of 102 young people (some were not students) participated in the study (52 females; M age = 22.80 years, SD = 3.62; age range 18–36 years). They were recruited by newspaper ads and flyers in Krakow, Poland. The same reward applied as in Experiment 1.
Materials and procedure

Stimuli were the same as in Experiment 1, with the exception that they were presented on a light gray background. In order to make this task more similar to the standard n-back task, we no longer used squares that separated the sequences of stimuli. Instead, one stream of 80 stimuli was presented serially to participants in each session. Each item was presented for 1,000 ms and then a mask was shown for another 1,000 ms. There were 16 targets in one session, 4 per each n-back condition (i.e., 1-, 2-, 3-, 4-back), placed randomly in the stream of stimuli with regard to n. The neighboring stimulus repetitions could not overlap. All 16 targets differed from one another, and none of the remaining 64 items could match any other stimulus in the whole session. Each session started with a fixation point presented for 1,000 ms. There was one training session and eight experimental sessions, each followed by a short break. The participants were instructed to remember the four most recent numbers and to press the space bar only when they detected stimulus repetition one, two, three, or four items ago, and to withhold responses in any other case. Correct responses had to occur within 250 ms of a target appearing on the screen and within 250 ms of a respective mask disappearing from the screen. All remaining responses were treated as errors.

In Experiments 2–6, we used two tests of fluid reasoning. A novel analogy test (Orzechowski & Chuderski, 2007) included figural analogies in the form “A is to B as C is to X,” where A, B, and C are types of figures; A is related to B according to two, three, four, or five latent rules (e.g., symmetry, rotation, change in size, color, thickness, number of objects); and X is an empty space. The task is to choose one figure from four that relates to Figure C as B relates to A. Before the WM task, each participant was given Raven's Advanced Progressive Matrices and the figural analogy test. The reasoning score (Gf) was calculated as the mean of z scores of the total number of correctly solved items within 60 min in Raven's test and the total number of correctly solved items within 60 min in the analogy test.
Results

Raven's test scores ranged from 2 to 36 (M = 22.14, SD = 8.35). Analogy test scores ranged from 3 to 36 (M = 23.97, SD = 8.08). The analogy test used herein was highly reliable (α = .86; based on a sample of 1,129 participants), which was comparable to the reliability of Raven's test (see Experiment 1). Both Gf test scores correlated strongly (r = .85, p < .001).

Reliabilities of the n-back task were α1 = .42, α2 = .54, α3 = .60, and α4 = .60. These values were unsatisfactory. The hit rates were again corrected by subtracting the individual rates of false alarms to non-repeated stimuli, which this time were relatively high (M = 0.15, SD = 0.15), from each hit rate. Again, there was a significant main effect of n factor, F(3, 303) = 125.46, η2 = .55, p < .001.

All correlations among hit rates were significant (rs > .50, ps < .001). The correlations between Gf and the 1-back (r = .50), 2-back (r = .47), as well as the 3-back (r = .32) were significant (ps < .001), while the correlation between Gf and the 4-back was not (r = .10, p = .318). The correlation coefficient for the 1-back was significantly higher than the r value regarding the 4-back, t(100) = 4.40, p < .001.
Discussion

The increase in time allowed for responding resulted in higher levels of accuracy, comparable to other n-back task studies. Experiment 2 replicated and extended the results of Experiment 1. The correlation coefficients between hit rates and reasoning were the highest for the 1- and 2-back targets and then significantly decreased as the value of n increased. This result is consistent with our hypothesis suggesting that only attentional capacity, including a very limited number of items, contributed to fluid reasoning. Performance under the largest WM load, which was presumably based on retrievals from WM area outside the focus, was not related to reasoning.

It is interesting that even the 1-back hit rate correlated with reasoning as strongly as did the 2-back rate, though the 1-back condition imposed a relatively low load on WM. According to our assumptions, maintaining the 1-back items required attentional capacity of two items (i.e., the current and 1-back items), so assuming a mean capacity equal to three or four items, most participants should have been able to fulfill such a condition within their foci of attention, and it should not have yielded particularly strong correlations with reasoning. However, maybe in the n-back task, due to its dynamic nature and the compound stimuli we used, some participants were not able to fully use their capacity, and they attentionally processed a smaller number of items than they would have processed in a less dynamic task. For this reason, the 1-back condition might also appropriately differentiate between participants who did and did not use the focus of attention to maintain the 1-back targets.

An unexpected outcome of Experiment 2 was the low reliability of the hit rates. However, the rates that yielded the weakest correlations with reasoning (i.e., the 3- and 4-back), had higher reliability than more strongly correlating rates (i.e., 1- and 2-back), and the former approached the reliability level accepted in psychometric studies (both αs = .60). So, it is unlikely that the observed differences in the correlation strength between hit rates and reasoning were caused by differences in reliabilities of the former. On the contrary, the differences in reliability might have attenuated the observed effect, which would be even stronger if the reliabilities of all hit rates were similar.
Experiment 3



Experiment 3 aimed to extend the results regarding the 4-back condition onto the 5-back condition. Moreover, having more conditions and a larger sample, we intended to test our aforementioned interpretations with structural equation modeling (SEM). We aimed to introduce one latent variable representing the hit rates at small n positions, assumed to fall within the focus of attention, and the other variable representing hit rates at large n positions, presumably reflecting encoding outside the focus, and to relate both these variables to the variable reflecting reasoning. We also wanted to compare such a model with a one-factor model, including one WM variable loaded by all hit rates.
Method
Participants

Participants were recruited via publicly accessible social networking websites. A total of 136 people participated (89 females; M age = 23.2 years, SD = 4.4; range 17–44). Each participant was paid the equivalent of about €6 in Polish zloty.
Materials and procedure

We used the same stimuli as in Experiments 1 and 2. This time a stream of 75 stimuli was presented serially to participants in each session. Each item was presented for 1,800 ms and was followed by a mask that was shown for 1,000 ms. There were 15 targets in each session, three per each n-back condition (i.e., 1-, 2-, 3-, 4-, and 5-back), placed randomly in the stream of stimuli with regard to n. There was one training session and five experimental sessions, each followed by a short break. Also, in this and subsequent experiments, a “warm-up” computerized task (dissimilar from the n-back task) was applied for several minutes, allowing the participants to get familiar with a computerized procedure. All other procedural details and the calculation of the dependent variables were identical to Experiment 2. In Experiment 3, 20 min were allowed for solving Raven's test, and 15 min were given for the figural analogy test, in order to shorten the duration of the experiment.

For SEM computations, we used Statistica software (Version 9) with the maximum-likelihood estimation. The goodness of fit of the present and subsequent models was evaluated with three measures: chi-square value divided by the number of degrees of freedom (χ2/df), Bentler's comparative fit index (CFI), and the root-mean-square error of approximation (RMSEA). We adopted the following commonly postulated criteria of the acceptable fit of models: χ2/df should not exceed 2.0, CFI should be higher than the value of .92, and RMSEA should not surpass the value of .08.
Results

Raven's test scores ranged from 2 to 29 (M = 18.75, SD = 4.85). Analogy test scores ranged from 3 to 31 (M = 17.63, SD = 5.23). Reliabilities of the n-back task were α1 = .52, α2 = .50, α3 = .46, α4 = .63, and α5 = .63, and they were still unsatisfactory. The hit rates were corrected by subtracting the (relatively low) individual rates of false alarms to non-repeated stimuli (M = 0.07, SD = 0.02). Again, the main effect of n factor was found, F(4, 540) = 221.44, η2 = .62, p < .001.

The significant (all ps < .02) correlations between the reasoning score and hit rates regarded the 1-back (r = .21), 2-back (r = .26), and 3-back (r = .28) hit rates, while the 4-back (r = .16) and 5-back (r = .13) conditions did not yield significant correlations (ps > .07). The (highest) correlation coefficient regarding the 3-back was significantly higher than the (lowest) r value of the 5-back condition, t(134) = 1.74, p = .042.

Correlations among hit rates and Gf test scores, used in SEM computations, are presented in Table 1.
TABLES AND FIGURES

Table 1. Correlation Matrix for Hit Rates and Gf Test Scores in Experiment 3

View larger image > in this page > in new window
> Download slide
They indicate that the longer the distance between conditions (e.g., the 1- and 5-back vs. the 4- and 5-back), the weaker the correlations between them. Exploratory factor analysis (varimax rotation) extracted two factors, one (eigenvalue = 1.46) loading the 1-, 2-, and (moderately) 3-back hit rates, and the other (eigenvalue = 2.16) loading the 3-, 4-, and 5-back hit rates.


The above data suggest that two latent variables may underlie performance in the n-back task in Experiment 3. One variable may reflect the effectiveness of the focus of attention, and it should be loaded by the 1-, 2-, and 3-back hit rates, while the other variable may represent the accuracy of retrievals from the activated part of LTM, and it should be loaded by the 3-, 4-, and 5-back hit rates. The former variable was expected to strongly correlate with the reasoning variable, which represented variance shared by Raven's and the figural analogy tests, while the latter variable may not be a significant predictor of reasoning. The overlap regarding the 3-back hit rate reflected individual differences in attentional capacity (Cowan, 2001). We assumed that some more capacious participants would be able to fulfill 3-back targets within their foci of attention, while others would not be able to do that, and by linking the 3-back hit rate to both variables, we were able to account for this fact. Such a model was estimated and is presented in Figure 1.
TABLES AND FIGURES

Figure 1. The structural equation model for Experiment 3, relating the focus of attention and activated long-term memory (LTM) exogenous latent variables to the fluid reasoning endogenous latent variable. Boxes represent manifest variables (see explanation in the text). Large ovals represent latent variables, while the small oval represents a disturbance term. Values between ovals and boxes represent relevant standardized factor loadings (all ps < .01). Values between ovals represent path coefficients among latent variables: The solid lines represent ps < .001, while the dashed line represents p > .05.
View larger image > in this page > in new window
> Download slide


The model's fit was excellent (N = 136, df = 10, χ2/df = 0.96, CFI = 1.0, RMSEA = .00). Both latent variables moderately correlated (r = .55), which is understandable as all their manifest variables reflected one and the same task. The variable representing the focus of attention explained 15.2% of variance in reasoning, while the other variable did not account for any Gf variance. Elimination of the path between activated LTM and reasoning did not influence the model's fit (Δdf = 1, Δχ2 = 0.08). An alternative model, which included one latent variable loaded by all five hit rates, was unacceptable (N = 136, df = 13, χ2/df = 2.56, CFI = .923, RMSEA = .122) and definitely worse than the two-factor model (Δdf = 3, Δχ2 = 23.68).
Discussion

The presented data again indicated that the n-back task is a good predictor of fluid reasoning only when n does not exceed three, and these data are consistent with the interpretation that only the capacity of severely limited focal attention contributes to reasoning. This time, the 1-back hit rate correlated more weakly with Gf than did the 2- and 3-back rates (though such a difference was not significant). Prolonging the trials allows us to reject the hypothesis that assumed that participants did not hold 4- or 5-back items in their WM due to time pressure. Contrary to the studies showing an important role of LTM in fluid ability (e.g., Mogle, Lovett, Stawski, & Sliwinski, 2008; Unsworth, Spillers, & Brewer, 2010), our study did not confirm any link between the activated part of LTM (related to performance on 4- and 5-back targets) and reasoning.

Again, although the reliabilities of hit rates were relatively low, the rates with the highest relative reliability yielded the lowest correlations with reasoning. This fact, as in Experiment 2, makes unlikely any interpretation suggesting that the observed effect is related to differences in reliability. One possible cause of the low reliabilities observed in Experiments 2 and 3 could be the fact that the processing in the inclusion version of n-back task is relatively weakly constrained, meaning that participants had to track more than one n-back position at once and their relevant coping strategies may have differed (in Experiment 1, due to only three n-back conditions and the fast presentation rate, this might not have been such a problem). If so, then the task's exclusion version, which requires tracking only one n-back position at a time, should give more reliable measures.
Experiment 4



In Experiment 4, we switched the procedure to the exclusion version of the n-back task. In the inclusion version, participants are required to simultaneously focus their attention on a few items. One question is whether we can also observe the effect of the n factor on the strength of WM–Gf correlation when participants' attention is focused on one prespecified value of n. Surely, items at more recent positions than the n-back have to be somehow maintained before they become n-back items. So, in the exclusion version, it can be expected that even though participants can allocate their attention to only one (n-back) item, increasing the value of n would still reduce the chance that this item would be maintained in the focus of attention. Thus, a decreasing strength of correlation between reasoning and hit rates in the function of n should be observed. In order to test that expectation we examined the 2-, 3-, and 4-back conditions separately, and we informed participants about which condition was currently being tested.

Another aim of this experiment was to examine responses to lures, which were introduced at the 1- and 5-back positions in all conditions. We asked if false alarm rates related to these lures would be significantly correlated with n-back performance in target trials and with reasoning.

It should be noted that in the exclusion version of the n-back task, the hit and false alarm rates of an individual are influenced not only by a bias related to guessing whether an item was or was not repeated (as indicated by rates of false alarms to non-repeated stimuli) but also by a bias related to whether individuals are more likely to interpret detected stimulus repetitions as targets (matching n) or lures (not matching n). However, in a particular condition of the n-back task paradigm, it is not possible to directly estimate such an individual bias (e.g., with use of the β parameter of signal detection theory) nor to derive a bias-free index of discriminability between lures and targets (e.g., d' parameter), because, by definition, targets and lures have to be placed at disjoint n positions in a stimuli sequence, and as the n effect on response accuracy is substantial, one cannot directly compare commission and omission errors at unequal n positions. For example, equal rates of false alarms for sub-n-back lures and hits for n-back targets would not indicate zero discriminability between the former and the latter, because a hypothetical hit rate at sub-n-back position would be higher than that at n-back, and thus the discriminability would in fact be positive (the reverse would be true in case of supra-n-back lures). Nonetheless, some approximate estimation of β and d' indices (as recommended by reviewers of a previous version of this article) seemed to be necessary in regard to our exclusion studies. So, regardless of the aforementioned objections, we undertook the following procedure in accounting for biases toward targets/lures.

First, in all subsequent experiments, we verified for a given n condition if a mean discriminability of lures and targets, both located at the same position, is satisfactory, because poorly discriminable lures most probably would not sufficiently engage control processes, and the corresponding false alarm rate would not be a valid indicator of executive control. So, we computed d' indices using mean hit and false alarm rates from different groups/experiments. Although the latter procedure did not allow us to compute individual values of d', on the assumption that the n-back procedure was relatively similar among consecutive experiments and it yielded comparable hit rates for the analogous conditions (see Appendix), d' indices computed in such a way should approximate the true mean discriminability at particular n positions. We were interested in whether a mean d' level was substantially above zero, indicating that participants were really able to identify lures as lures (and then the differences in suppressing responses to these lures can be validly interpreted), or if it was close to zero, suggesting that they messed up targets and lures.

Second, we computed individual d' values for each n-back condition regarding targets, using false alarm rates from the same condition. Although these estimates deviated from the true values of discriminability for the aforementioned reasons, they more or less approximated the individual bias-free level of performance at various ns. As we were not interested in absolute values of these estimates but used them only in correlational analyses (i.e., in relation to Gf), such an approximation seemed acceptable.

Finally, using the same hit and false alarm rates as in individual d' calculation, we also approximated decisional biases in each n-back condition. We did so by dividing the mean false alarm rate observed in that condition by the mean sum of rates of false alarms and omissions (i.e., the latter equal to one minus hit rate) in that condition, no matter what lure positions were applied. We aimed to test whether the expected advantage of intelligent participants over less intelligent ones in some conditions of the exclusion n-back task would or would not emerge due to their more liberal or more conservative bias. As we were again interested only in relative differences in the values of β regarding intelligence level, such an approximation also seemed acceptable. In addition to analyses regarding the d' and β parameters, we separately analyzed data on hit and false alarm rates.
Method
Participants

A total of 135 participants, who were tested in Experiment 3, were examined with the exclusion version immediately after completing the inclusion one (one remaining male person resigned).
Materials and procedure

Stimuli were the same as in Experiment 3. A stream of 84 stimuli was presented serially in each session. Three sessions were used for each n condition (2-, 3-, and 4-back). There were 10 targets in each session. Also, three 1-back repetitions and three 5-back repetitions were included in each session, acting as lures. No other stimuli could be repeated within the same session. Participants were informed of how many items back the targets would appear in a given session, and they were instructed not to respond to the 1- and 5-back lures. Also, a proper instruction was presented at the bottom of the screen for the entire duration of a session (e.g., in the 3-back, “respond only to item repetitions separated by exactly two other items”). We expected that rejecting lures would involve control to some extent, because in the preceding inclusion task participants had learned to respond to the 1- and 5-back targets. Half the participants fulfilled the task in the following order: the 2-, 3-, and 4-back. The other half of participants encountered the reversed order of sessions. All other procedural details and dependent variables calculations were identical to Experiment 3.
Results

The reasoning scores from Experiment 3 were adopted. Reliabilities of the hit rates equaled α2 = .88, α3 = .86, and α4 = .84. They were substantially higher than in the two preceding experiments and were very satisfactory. Reliabilities of the false alarm rates were α1 = .84 and α5 = .65. The hit rates were corrected by subtracting the individual rates of false alarms to non-repeated stimuli in the respective n-back conditions. False alarm rates were corrected accordingly using the individual rates of false alarms to non-repeated stimuli averaged over all conditions (M = 0.07, SD = 0.02). The main effect of the n factor on hit rate was significant, F(4, 540) = 221.44, η2 = .62, p < .001, as was the n effect on false alarm rate regarding the 1-back versus 5-back lures, F(4, 540) = 6.23, η2 = .04, p = .014. The latter effect indicated that the 5-back false alarm rate was higher than the 1-back rate.

In order to calculate the mean d' values for the 1- and 5-back positions, we used the 1- and 5-back hit rates from Experiment 3. The d' value for the former position equaled d' = 2.39, while such a value for the latter position was only d' = 0.66. So, the 1-back lures were easily discriminable, most probably because they constituted direct repetitions of stimuli and their inadequate position could be clearly identified, while the 5-back ones could not be discriminated so well, and most probably the more 4-back items one could recognize, the more 5-back alarms the participant had committed due to the greater number of detected 5-back repetitions.

The latter conclusion was supported by inspecting the correlations among hit/false alarm rates and scores on Gf tests, presented in Table 2.
TABLES AND FIGURES

Table 2. Correlation Matrix for Hit/False Alarm Rates and Gf Test Scores in Experiment 4

View larger image > in this page > in new window
> Download slide
It shows that the 3- and 4-back hit rates positively correlated with the 5-back false alarm rate, while the 1-back false alarm rate correlated negatively with the 2-back hit rate. Also, all three hit rates strongly correlated, and there was also a significant correlation between the 1- and 5-back false alarm rates. Analysis of correlations between Gf and the individual d' values for 2-, 3-, and 4-back positions yielded significant rs (all ps < .01); however the 2-back (r = .44) correlation was significantly stronger than the 3-back (r = .22) and the 4-back (r = .27), with t(133) = 2.08, p = .02, for the latter comparison. All individual d' values moderately correlated (.34 < rs < .44, ps < .001). The analysis of β parameters for the 2-, 3-, and 4-back positions did not indicate any significant correlation with Gf (all rs < |.095|, ps > .27).


All correlations between the reasoning score and hit rates in all n-back conditions were significant and equaled r = .36, p < .001; r = .25, p = .004; and r = .18, p = .038, in the 2-, 3, and 4-back conditions, respectively. The correlation regarding the 2-back was significantly stronger than the 4-back correlation, t(133) = 2.23, p = .014. The correlation between Gf and the 1-back false alarm rate was significant (r = –.22, p = .011), while the 5-back correlation was not (r = –.05, p = .549).

Exploratory factor analysis (varimax rotation) applied to the hit and false alarm rates extracted two factors, one (eigenvalue = 2.07) highly loaded three hit rates, while the other (eigenvalue = 1.44) highly loaded the false alarm rates. So, accordingly, we tested an SEM model in which all hit rates loaded onto one latent variable, both false alarm rates loaded onto another latent variable, and these two variables correlated and predicted the reasoning variable. However, such a model yielded an unacceptable fit (N = 135, df = 11, χ2/df = 4.21, CFI = .838, RMSEA = .159). Second, taking into account the results of that d' analysis, which showed that it was difficult for participants to distinguish the 5-back lures and 4-back items, we altered the model by allowing the 5-back false alarm rate to load onto both exogenous variables. This helped only a little, and the modified model also fit poorly (N = 135, df = 10, χ2/df = 3.58, CFI = .882, RMSEA = .140).

Finally, we tested an SEM model including three latent variables, one aimed to reflect the focus of attention, the second probably representing the activated part of LTM (i.e., analogously to the model estimated in Experiment 3), and the last variable intended to account for executive control. We allowed the 2- and 3-back hit rates to load onto the focus of attention variable, while the 3-back and 4-back hit rates and the 5-back false alarm rate loaded onto the activated part of LTM. Both false alarm rates loaded onto the executive control variable. All these latent variables were allowed to correlate, and each was linked by a directed path to the reasoning variable. The model's fit was excellent (N = 135, df = 7, χ2/df = 0.58, CFI = 1.0, RMSEA < .001). However, activated LTM again predicted a negligible amount of reasoning variance (r = .07, p = .613), and elimination of this path did not influence the fit (Δdf = 1, Δχ2 = 0.25). The resulting model is presented in Figure 2.
TABLES AND FIGURES

Figure 2. The structural equation model for Experiment 4, relating the focus of attention and executive control exogenous latent variables to the fluid reasoning endogenous latent variable. Boxes represent manifest variables (see explanation in the text). Large ovals represent latent variables, while the small oval represents a disturbance term. Values between ovals and boxes represent relevant standardized factor loadings (all ps < .01). Values between ovals represent path coefficients among latent variables: The solid lines represent ps < .001, while the dashed lines represent ps > .05. LTM = long-term memory.
View larger image > in this page > in new window
> Download slide
The one-factor model definitely fit worse than this model (Δdf = 3, Δχ2 = 32.12). The focus of attention variable moderately correlated with the activated LTM (r = .38, p < .001) and executive control (r = .32, p = .016) variables. The path between the focus of attention and reasoning variables was highly significant (r = .40, p = .001), while the path between control and reasoning was only marginally significant (r = .22, p = .087).

Discussion

Experiment 4 yielded three new results. First, also in the exclusion condition, the correlation between reasoning and hit rates weakened with increasing n, which again favors an interpretation that reasoning ability is related to the limited capacity of the focus of attention.

Second, the d' parameter analysis indicated that the 5-back lures were difficult to identify as lures and they were often treated as targets. As such, they did not constitute effective distractors and the lack of significant correlation between their rejection rate and reasoning cannot be meaningfully interpreted.

Finally, the 1-back lures, contrary to the 5-back ones, were easily identifiable and seemed to work as effective distractors. Most probably, as direct repetitions, their exact n-back position could be easily assessed. This view is supported by the fact that in Experiments 2 and 3 the 1-back items were almost perfectly detected (M = 0.91 on average). However, even though it seems that participants should have been able to correctly recognize almost all 1-back lures as lures, they committed a surprisingly high rate of false alarms to these lures (M = 0.18) and varied substantially in how they rejected them (SD = 0.20; range 0.0–1.0). Such a result suggests that identification of an item as a 1-back lure was not always sufficient to reject it. The role of executive control might be in helping to suppress reactions to lures, which otherwise would be made regardless of a participant's knowledge that they are indeed lures. The involvement of executive control would be analogous to the case of the Stroop task, when, regardless of participants' knowledge that the task consists of naming colors, the control has to prevent them from reading words. How well participants rejected such lures did correlate with reasoning, which suggests that such a kind of control may be another contributor to fluid intelligence. The SEM analysis suggested that in contributing to reasoning the variable reflecting control is independent from the variance reflecting capacity. However, the former factor accounted for three times less Gf variance as the latter.
Experiment 5



In this experiment, we applied the between-subjects design with regard to the n factor (i.e., the 2-back vs. 3-back groups). We used n − 1 lures (i.e., 1-back and 2-back, respectively) and n + 1 lures (i.e., 3-back and 4-back, respectively), in order to test if the 2-, 3-, and 4-back lures could also yield significant correlations between false alarms and reasoning as did the 1-back lures, or if they could not yield it, similar to 5-back lures. Finally, we used figural stimuli instead of numerical ones. The former are more difficult to maintain and should substantially reduce chunking and rehearsal, thus amplifying correlations between the n-back task performance and reasoning.

The experimental procedure was further extended by introducing a new task modeled on Luck and Vogel's (1997) two-array comparison task, which is considered (Cowan et al., 2006; Rouder, Morey, Morey, & Cowan, 2011) to allow for reliable estimation of individual focus of attention capacity. The original task required memorizing an array of items. Then, after a retention interval, the array was repeated, but there was 50% chance that one of the items would change. The task was to indicate whether the item had changed or not. In Cowan et al.'s (2006) version of this task, one of the items in the second array was marked by a cue indicating that, if any of the items had changed, it was the marked one.

The formula, which estimates the sheer capacity of the focus of attention as the proportion of hits (H, correct responses for arrays with one item changed) and the proportion of false alarms (FA, incorrect responses for unchanged arrays), in the cued version of a two-array task, was proposed by Cowan (2001). Accordingly, the capacity of the focus is estimated to be k items (out of N items of a memory load), on the assumption that a participant produces a correct hit or avoids a false alarm only if a cued item is transferred to his or her focus (with the k/N chance). If a non-transferred item is cued, then a participant guesses the answer. Consequently, the following formula evaluates the capacity of the focus of attention: k = N × (H – FA).

With the use of the two-array comparison task and the estimation of k, we aimed to test our assumption that accuracy in low n conditions of the n-back task is related to individual differences in capacity of the focus of attention. We assumed that the 2-back condition involves maintenance of a minimum of three items in the focus: a currently presented stimulus, a 1-back one (which is needed for a next trial), and a 2-back one (which has to be compared with a current stimulus). Similarly, in the 3-back condition a minimum of four items have to be maintained in the focus in order to provide high accuracy. Of course, participants with insufficient capacity may still respond correctly on some occasions using their activated LTM. We hypothesized that in the 2-back condition participants showing k values of three and more would display higher accuracy than ones presenting ks below three, while in the 3-back condition only participants possessing k values of four or more would surpass less capacious ones.
Method
Participants

A total of 275 people, who were recruited via publicly accessible social networking websites, participated (166 females; M age = 22.5 years, SD = 3.6; range 17–40). Each participant was paid the equivalent of €7 in Polish zloty. The participants were assigned to two groups (the 2-back group had 135 people and the 3-back group had 138 people) according to the order in which they entered the laboratory. Data from two participants (one in each group) were excluded because of enormously high error rates (M = 0.70 and M = 0.49), greatly exceeding the error rates of all other participants, which suggested that these two participants did not follow the task instructions.
Materials and procedure

In the n-back task, stimuli were 16 simple black figures (e.g., a square, a circle, a rhombus, an arrow, a cross), each approximately 2.5 × 2.5 cm in size, presented for 1,500 ms plus a 300-ms mask. A total of 88 stimuli were presented serially to participants in each session. Four sessions were used in each group, preceded by some training. Each session that was completed by the 2-back group included eight 2-back targets, four 1-back lures, and four 3-back lures. In the 3-back group, eight 3-back targets, four 2-back lures, and four 4-back lures were included in each session. No other stimuli could be repeated within the few most recent items. Participants were instructed to respond to either 2-back or 3-back repetitions (depending on the group) and to suppress responses to all other repetitions. As only one n condition was applied to each participant, there were no hints presented during trials. Calculation of the dependent variables was identical to that in previous experiments.

Each of the 90 trials of the two-array comparison task consisted of a virtual, four by four array filled with a few stimuli (i.e., only some cells in the array were filled). The stimuli were 10 Greek symbols (e.g., α, β, χ), each 2 × 2 cm in size. The number of stimuli within the array varied from five to seven items. The array was presented for the time equal to the number of its items multiplied by 400 ms and then followed by a black square mask of the same size as the array, presented for 1,200 ms. In a random 50% of trials, the second array was identical to the first one, while in the remaining trials both differed by exactly one item at one location. If they differed, the new item was highlighted by a square red border. If they were identical, a random item was highlighted. The task was to press one of two response keys depending on whether the highlighted item differed or not in the two arrays. The second array was shown until a response was given or 4 s elapsed. The trials were self-paced. The dependent variable was a mean of k values calculated for each set size.

In Experiment 5, we allowed 40 min for Raven's Progressive Matrices, and 30 min were given for the figural analogy test.
Results

The Raven test scores ranged from 1 to 36 (M = 22.77, SD = 6.40). Analogy test scores ranged from 10 to 36 (M = 25.32, SD = 5.45). Reliabilities of the hit rates were α2 = .85 and α3 = .78. Reliabilities of the false alarm rates were α1 = .92, α2 = .68, α3 = .76, and α4 = .64. All these values ranged from good to acceptable. The hit and false alarm rates were corrected by subtracting the individual rates of false alarms to non-repeated stimuli, which were low (M = 0.06, SD = 0.06).

There was the main effect of the n factor in both targets and lures, F(1, 271) = 53.80, η2 = .17, p < .001, and F(1, 271) = 35.32, η2 = .12, p < .001, respectively. The former effect reflected a higher hit rate in the 2-back than in the 3-back condition, while the latter effect indicated that the 2- and 4-back false alarm rates observed in the 3-back group were significantly higher than the 1- and 3-back false alarm rates seen in the 2-back group (M = 0.41 vs. M = 0.20, respectively). We also calculated the d' indices using the 2- and 3-back hit rates from the present experiment and the 1- and 4-back hit rates from Experiment 3. The resulting values were d' = 2.39, d' = 0.78, d' = 0.87, and d' = 0.51, for the 1-, 2-, 3-, and 4-back conditions, respectively.

The correlational matrix for lure and hit rates is presented in Table 3.
TABLES AND FIGURES

Table 3. Correlation Matrix for Hit/False Alarm Rates and Gf Test Scores in Experiment 5

View larger image > in this page > in new window
> Download slide
Matching the analysis of d' values, it shows significant positive correlations between the hit rates and the 2-, 3-, and 4-back false alarm rates but not between the former and the 1-back false alarm rate. Both d' values and those correlations indicate that participants imperfectly distinguished 2-, 3-, and 4-back lures from targets. The individual d' value significantly correlated with reasoning both in the 2-back group (r = .51, p < .001) and in the 3-back group (r = .26, p = .002), which differed significantly, t(271) = 3.20, p < .001. The correlations between the β parameter and reasoning were significant neither in the 2-back group (r = –.04, p = .652) nor in the 3-back group (r = .03, p = .705).


The 2-back hit rate correlated with reasoning score significantly more strongly (r = .50, p < .001) than did the 3-back hit rate (r = .25, p = .004), t(271) = 3.16, p < .001. In the 2-back group, the 1- and 3-back false alarm rates significantly correlated with reasoning (r = –.22 and r = –.20, respectively, both ps < .03). On the contrary, in the 3-back group, correlations of the 2- and 4-back false alarms with Gf were substantially weaker and not significant (r = –.05 and r = .04, respectively, both ps > .5).

The mean k value equaled M = 3.25 (SD = 1.36, range 0.0–5.78). In order to present the differences in n-back performance related to the k estimate, we divided participants into five groups. The k = 1 group (n = 38) included participants with k values lower than 1.5, who were assumed to be able to track on average only the current item. The k = 2 group (n = 31) included participants in the range 1.5–2.5, who were assumed to be able to hold on average two items in their foci of attention. The k = 3 (n = 68) and k = 4 (n = 88) groups were formed accordingly. The k = 5 (n = 48) group included participants possessing k values of 4.5 or higher. The means for all k groups and both n-back conditions are presented in Figure 3.
TABLES AND FIGURES

Figure 3. The error-corrected hit rates in five groups differing in the estimated capacity of the focus of attention (k values), in the 2-back (solid line) and the 3-back (dashed line) conditions, in Experiment 5. Error bars represent 1 SE of the mean.
View larger image > in this page > in new window
> Download slide
Subsequently, hit rates in the n-back task were submitted to ANOVA, with k (1–5) and n-back (2 and 3) groups as factors. The effect of the k group was highly significant, F(4, 263) = 13.29, η2 = .17, p < .001, indicating that accuracy in both tasks were strongly related. Most important, the interaction between k and n group factors was also significant, F(4, 263) = 4.43, η2 = .04, p = .048. It showed that in the 2-back group, there was no contrast between the k = 1 and k = 2 groups (F = 0.05), while there was a highly significant contrast between the k = 2 and k = 3 groups, F(1, 263) = 7.06, p = .008 (further contrast between the k = 3 and k = 4 groups was significant, F[1, 263] = 4.32, p = .038, and the one between the k = 4 and k = 5 groups was close to significance, F[1, 263] = 3.48, p = .063). On the contrary, in the 3-back group, the k = 2 and k = 3 groups did not differ significantly (F = 0.03), while the k = 3 and k = 4 groups did, F(1, 263) = 6.63, p = .011 (the k = 4 and k = 5 groups contrast was not significant, F = 0.25). Contrary to the hit rates, no effect of the k group on the false alarm rates has been found (F = 0.83).


Data regarding the 2-back group, indicating good predictability of reasoning by hit rates as well as sufficient identifiability of lures, were used in order to calculate another SEM model, relating the attentional capacity and executive control variables to the reasoning variable. We aimed to replicate the results of SEM analysis done in Experiment 4. In the present model, the capacity variable was loaded by the 2-back hit rates and also by another measure of capacity, namely, the value of k. The control variable was loaded by the 1- and 3-back lures. These assumptions were supported by exploratory factor analysis (varimax rotation), which showed one factor (eigenvalue = 1.57) yielding high loadings (>.82) on the 2-back hit rates and k values but negligible loadings (<|.25|) on both false alarm rates, and another factor (eigenvalue = 1.25) yielding high loadings (>.80) on false alarms but negligible loadings (<|.05|) on the hit rate and the value of k. According to the good-fitting model (N = 135, df = 7, χ2/df = 1.76; CFI = .969, RMSEA = .071), presented in Figure 4,
TABLES AND FIGURES

Figure 4. The structural equation model for the 2-back group of Experiment 5, relating the focus of attention and executive control exogenous latent variables to the fluid reasoning endogenous latent variable. Boxes represent manifest variables (see explanation in the text). Large ovals represent latent variables, while the small oval represents a disturbance term. Values between ovals and boxes represent relevant standardized factor loadings (all ps < .01). Values between ovals represent path coefficients among latent variables: The solid lines represent ps < .001, while the dashed line represents p > .05. Note that because of the negative variance found in the initial model, the error parameter for the 3-back false alarm rate was fixed at a value of 1 minus its reliability.
View larger image > in this page > in new window
> Download slide
attentional capacity and executive control did not significantly correlate (r = .13, p = .291), and the former variable explained 62.4% of variance in reasoning (r = .79, p < .001), while the latter accounted for 13.7% of that variance (r = .37, p < .001). The correlation between capacity and control could be eliminated without significant loss of fit (Δdf = 1, Δχ2 = 1.07). We compared the above model with the one-factor model, which included a latent variable on which all four manifest variables loaded. Such a model appeared to be completely unacceptable (CFI = .452, RMSEA = .296).

Discussion

In the figural version of the n-back task, the correlation between accuracy in the task and the reasoning score also appeared to be significantly stronger for a lower value of n than for its higher value. This result clearly replicated the data regarding the numerical version of the task, observed in Experiments 1–4. Additionally, the analysis of individual k values, estimated with the two-array comparison task, presumably reflecting how many items could be held within the focus of attention, indicates that such a capacity substantially varied in our sample, from one (if we assume that several zero values of k mostly reflected little involvement in fulfilling the task) up to almost six items. The analysis of relation between performance on both WM tasks provided a high level of correspondence between the assumed capacity requirements of the 2- versus 3-back conditions and the special advantages of the k = 3 and k = 4 groups, respectively. This result supports such an interpretation of processing requirements of the n-back task, suggesting that the capacity of three items is needed for maintaining the 2-back targets, while the capacity of four items is needed to attentionally process the 3-back targets.

False alarms in the 1- and 3-back lure conditions significantly contributed to reasoning, while 2- and 4-back ones did not. Together with results regarding the 5-back lures in Experiment 4, these data suggest that in order to effectively reject lures, participants have first to correctly identify them as lures and not as targets. In the 2-back group, as indicated by much lower false alarm rates in comparison to the 3-back group, differentiating lures from targets was relatively effective. As noticed when discussing Experiment 4, the 1-back lures were probably easily identified because they were direct repetitions of stimuli. The 3-back lures could have been relatively well identified because their position directly followed a position being tested for targets. In such trials, after a participant decided with high accuracy that a 2-back item was not a target, she or he encountered as the next stimulus a repetition of the target item, which as such could not be a 2-back one. Because of the good distinctiveness of lures in the 2-back group, false alarm rates in this group most probably properly reflected the efficiency of executive control and thus correlated significantly with reasoning. The contribution of executive control to Gf was independent from the contribution of capacity. The former was much smaller than the latter, but it was still significant.

On the contrary, in the 3-back group, lures seemed to be often indistinguishable from targets, probably due to inferior access to the more remote n positions. Data regarding lures most probably confounded executive control (when lures were correctly identified as lures) with storage capacity (when they were taken as targets). As lures and targets yielded opposite correlations with reasoning (i.e., negative vs. positive ones), it resulted in close-to-zero correlations of respective false alarm rates with Gf. As was previously noticed, such data cannot be meaningfully interpreted.

Of course, all the above regularities were probabilistic, so sometimes even k = 1 or k = 2 participants could have been able to maintain the 2-back or 3-back items in the focus (e.g., because they did not update the 1-back items, their attention was temporarily spanned beyond the mean capacity, or some items were chunked). A methodological conclusion follows that the nature of contribution to reasoning of any assumed measure of executive control derived from a WM task depends on the load on WM that is implicated in the processing of corresponding items. Only a rejection rate of easily identifiable and thus prepotent distractors directly reflects the effectiveness of executive control. Maybe, in the context of the n-back task, including larger distances between targets and lures will make the former and the latter more distinguishable. Only then will false alarms for lures at larger n-back positions than the 1-back strongly correlate with reasoning. This prediction is the subject of the next study.
Experiment 6



This small-sample follow-up study had one aim: to test if 4-back lures, when placed at a larger distance from targets (i.e., the 2-back ones), would be more correctly rejected than the 4-back lures in Experiment 5, and, if so, whether their rejection rate would be a significant predictor of reasoning.
Method
Participants

A total of 75 people, who were recruited via publicly accessible social networking websites, participated (47 females; M age = 23.5 years, SD = 4.5; range 18–46). Each participant was paid the equivalent of €6 in Polish zloty.
Materials and procedure

Stimuli were identical to those in Experiment 5. A total of 80 stimuli were presented serially to participants in each of two sessions (preceded by training). Each session included four 2-back targets and twelve 4-back lures. The participants were warned that lures are frequent and that they should be carefully rejected, while 2-back items should be responded to. All other procedural details, including Gf test administration time and the calculation of dependent variables, were identical to those of Experiment 5.
Results

Raven's test scores ranged from 6 to 35 (M = 21.73, SD = 6.02). Analogy test scores ranged from 5 to 36 (M = 23.83, SD = 6.30). Both Gf test scores correlated significantly (r = .74, p < .001).

Cronbach's alpha equaled α2 = .64 in hit rates (note that only eight targets were used) and α4 = .90 in false alarms. Participants committed a false alarm rate almost two times lower than the analogous 4-back false alarm rate in Experiment 5 (see the Appendix). The hit and false alarm rates were corrected by subtracting the individual rates of false alarms to non-repeated stimuli, which in any case were very low (M = 0.04, SD = 0.02). The mean d' value for the 4-back position, which was calculated with use of the 4-back hit rate in Experiment 3, was d' = 0.99, which suggests acceptable discriminability. The correlation between reasoning and individual d' equaled r = .47, p < .001, while reasoning and β did not correlate significantly (r = –.04). The hit and false alarm rates did not correlate significantly (r = –.09, p = .402) either, and, most important, each rate significantly and substantially correlated with the reasoning score (r = .36, p = .002, r = –.37, p = .001, respectively).
Discussion

The results of Experiment 6 suggest that false alarm rates regarding some lures placed at positions more than 1-back, as observed in Experiments 4 and 5, were not proper measures of executive control, because these lures were not effective enough as distractors and consequently they did not yield significant correlations with reasoning. When lures and targets were placed on distinctive n positions, namely, ones differing by n = 2, the former predicted reasoning as well as did the latter. The present experiment seems to support our previous conclusions, with regard to the fact that some form of executive control, operationalized as the ability to reject distinctive distractors in WM, may be a genuine explanatory factor of fluid reasoning, and it contributes to intelligence independently from WM capacity, expressed as the number of items held in the focus of attention.
General Discussion


Summary of Results

The presented research tested the theories that integrate capacity and control approaches in order to explain the cognitive basis of fluid reasoning. Summing up the results, in Experiments 1–6, the indices of performance, which were assumed to reflect storage capacity within WM, explained a substantial part of variance in fluid reasoning. In Experiments 4–6, some of the indices of performance, which were intended to measure the efficiency of control over distraction within WM, also predicted a significant but relatively smaller amount of variance in reasoning ability. As much as three quarters of variance in reasoning could be explained by these two factors (i.e., 76.1% in Experiment 5). The modified n-back task has been confirmed as a valid measure of WMC.
Storage Capacity and Fluid Reasoning

Regarding the first question, namely, whether either the total capacity of WM or the capacity of a particular substructure of WM is related to fluid reasoning, the results consistently suggest the latter possibility. In all six experiments, hit rates in our WM task substantially correlated with reasoning when the load imposed on WM was relatively low (i.e., n did not exceed three), while the respective correlation was substantially decreased or even eliminated in cases of larger values of n (Experiments 1–5). Such a pattern of results was also supported by two SEM models (Experiments 3 and 4), which included a null path between reasoning and the latent variable reflecting response accuracy for items at high n-back positions, while the variable related to items at low n-back positions predicted a substantial amount of variance in reasoning.

Why is accuracy at small n values strongly correlated with reasoning, while at large ones it is not? In our view, the 2-back condition is the most sensitive to the substantial individual differences in attentional capacity that we observed in the results from the two-array comparison task. Comparison of 2-back performance among participants with different k values indicates that half of the sample, namely, people with capacity lower than the mean k equaling around three items, were most probably not able to span their attention onto 2-back targets on most occasions, as such spanning required at least the capacity of three items (i.e., holding a current stimulus and two preceding items). So, in coping with the 2-back repetitions, those people possibly had to rely on the activated part of LTM, and due to that fact their n-back accuracy was decreased in comparison to the other half of the sample, which had a capacity surpassing the mean. Moreover, if we assume that in our n-back task participants did not always use their maximum capacity (i.e., the one estimated by their value of k), but due to the dynamic nature of this task and its compound stimuli were often able to hold in the focus of attention a lower number of stimuli, then a strong correlation between reasoning and the 1-back performance (as observed in Experiment 2) is also coherent with the above explanation.

On the contrary, the hit rates regarding large n values seem to poorly capture the individual differences in attentional capacity. In the 3-back condition, only the capacity of at least four items allowed the 3-back targets to be kept within attention (as indicated by increased accuracy in the k = 4 and k = 5 groups), so the requirements of that condition seem to exceed the capacity of most of the sample, and the 3-back targets were mostly retrieved from activated LTM. As only a minority of participants contributed to the observed individual differences in the 3-back hit rate, its correlation with Gf was usually much weaker than the analogous correlation of the 2-back (see Experiments 1, 2, 4, and 5). Extrapolating that line of reasoning, in the 4-back and, especially, 5-back conditions, hardly anyone could maintain n-back items in the focus, so the respective hit rates did not reflect individual differences in the capacity of the focus of attention, and it should be no surprise that there were no significant correlations regarding these conditions and Gf (except for a weak correlation regarding the 4-back in Experiment 4). This occurred even though reliabilities regarding the 4- and 5-back targets were comparable to those at lower n positions.

If the above interpretations are valid, then the observation of relatively strong correlations of the 2-back hit rates with reasoning, and the lack of significant correlations in most cases of the 4- and 5-back hit rates, leads to the aforementioned explanation that differences in reasoning scores depend on differences in the very limited capacity of the focus of attention, while they are not related to differences in the accuracy of retrievals from outside the focus. Hence, the presented data seem to speak for a much more important contribution to fluid processing of information by the focus of attention (primary memory) than by the activated part of LTM (secondary memory).

This thesis contrasts with interpretations by Mogle et al. (2008), Unsworth and Engle (2006), and Unsworth et al. (2010), who concluded that fluid intelligence is also related to the efficacy of search within secondary memory (e.g., to the narrowing of a search set). However, the paradigms used in our study (the n-back and two-array comparison tasks) and the ones used in the cited studies (the complex span and free recall tasks) differed significantly, for example by using the procedure of either recognition or recall, respectively, or by allowing relatively short (the n-back) or long (the free recall task) duration of time for encoding stimuli. In fact, in complex and long-lasting recall tasks, the use of STM and LTM may be substantially interleaved, so an issue of what the complex span and free recall tasks really measure and what is a relation of LTM to reasoning surely needs further research.

Why is the focus of attention so important for fluid cognition, while WM outside the focus is not? Although the present study cannot give an answer to that question, it would be interesting to consider a few options. The first one concerns the possibility that access to information in the focus may be much better than access to activated LTM. The former may yield rich representations, for example, a complete set of bound features of an item (e.g., an item's identity and its n-back position in the n-back task), while the latter may only indicate the global familiarity of an item (e.g., showing if it was/was not encoded, but not at what n-back position). Such an explanation would be coherent with our data showing that more recent targets yielded higher accuracy than did less recent ones, and also with data on the low discriminability of the 4- and 5-back lures from comparable targets in Experiments 5 and 4, respectively. However, in Experiment 5, the 2-back lures were also poorly discriminated, though they should have been processed in the focus of attention. Moreover, the 4-back lures of Experiment 6, presumably processed out of the focus, were rejected at a high rate. It seems that information on a serial position associated with a particular item can also be retrieved from activated LTM, at least in some cases. So, these results suggest that there seems to be more to the story concerning the relation between reasoning and the subparts of WM than just the ease of access to these subparts.

Another possibility concerns the fact that some inherent characteristic of the focus of attention may make accessing the focus only relatively more efficient than accessing the activated LTM, but it may be crucial for reasoning. A good candidate for such a characteristic, as described in the introduction, would be the ability of the focus of attention to represent arbitrary, relational structures (Halford et al., 1998, 2007), for example, due to constructing, maintaining, and transforming the flexible, temporary bindings among elements of such structures (Hummel & Holyoak, 2003; Oberauer et al., 2007). Operating easily on such structures seems to be crucial for both deductive reasoning (e.g., integration of the premises into mental models; Johnson-Laird, 1999) and inductive thinking (e.g., mapping elements of source and target during analogy making; Hummel & Holyoak, 2003; Waltz et al., 1999).
Executive Control and Fluid Reasoning

As to the second question posed in the introduction, our results suggest that some form of control over distraction seems to be an important prerequisite of effective reasoning, as false alarm rates significantly correlated with scores on intelligence tests. However, this observation pertains only to lures that were sufficiently distinctive from targets. On the contrary, results regarding indiscriminable lures are not meaningful, as they most probably reflected mixed cases when participants identified lures as lures but failed to reject them and cases when participants did not identify them as lures and responded to them in the belief that they were proper targets. Due to the latter fact, weak positive correlations between respective false alarm and hit rates were observed in cases of lures yielding low values of d'.

When the analysis of false alarm data was narrowed only to properly discriminated lures, there was barely any relation between control and capacity, as those false alarm rates did not significantly correlate with adequate hit rates (if any correlations were significant, they were weak and negative). Also, the strength of Gf correlations regarding those properly identified lures hardly depended on the value of n: Correlation for the 1-back lures in Experiments 4 and 5 equaled r = –.22, for the 3-back (Experiment 5) it equaled r = –.20, and for the 4-back (Experiment 6) it was r = –.37. In light of these results, the contribution of control to reasoning seems to be relatively independent from the contribution of attentional capacity. This conclusion is further supported by two SEM models (Experiments 4 and 5), which both estimated relatively weak paths (r = .36 and r = .13, respectively) between control and capacity. Although these results cannot exclude some overlap between capacity and control aspects of WM in predicting fluid reasoning, the amount of this overlap seems to be substantially smaller than the previous research suggested (e.g., Cowan et al., 2006).

What is the role of executive control in coping with lures, if any? One possibility (suggested by a reviewer) could be that rejecting lures depends solely on retrieving the proper n-back position of a repeated stimulus: If the retrieved serial position equals n indicated in the task's instruction, a repetition is accepted, while if they mismatch, a (lure) repetition is rejected. Although hit and false alarm rates would rely on the same process (i.e., the better one retrieves an n-back position, the better one both accepts targets and rejects lures), because of individual differences in bias for guessing targets/lures in cases when the serial position cannot be retrieved, the correlation between both these measures could be artifactually decreased (as noted by another reviewer), leading to a (false) identification of two distinct factors (i.e., capacity and control) in SEM models.

However, two results suggest that false alarms produced in response to highly available lures were caused not only by improper assessment of their positions. First, as noted previously, the average rate of false alarms to the 1-back lures (Experiments 4 and 5) was surprisingly high (M = 0.18), given that the 1-back repetitions were highly available to participants (e.g., a mean hit rate in Experiments 2 and 3 was M = 0.91) as well as the fact that they were also highly discriminable from targets (in Experiments 4 and 5, d' = 2.39). Second, if the high rate of false alarms to those highly discriminable lures was caused by imperfect assessment of serial positions of repetitions, then such an invalid process should also influence the 2-back hit rates (i.e., participants would omit some targets due to retrieving their wrong positions). On the contrary, in our data, the 2-back hit rates in the exclusion version, which were highly dependent on proper assessment of serial positions, were comparable to the 2-back hit rates in the inclusion version, in which serial positions needn't be assessed at all (M = 0.75 vs. M = 0.76, in Experiments 1–3 and Experiments 4–6, respectively). In other words, introducing the 1-back lures did not decrease the rate of acceptance of the 2-back targets. If participants knew that the 1-back repetitions were lures and the 2-back repetitions were targets (as suggested by these two results), why did they produce so many false alarms to the 1-back lures?

In our interpretation, within the go/no-go method that was applied, participants displayed a reflexive tendency to respond to whichever repetition they detected, and on some occasions this tendency could not be overridden even if they assessed that it was related to an improper n-back position. Suppressing such a tendency would be a kind of executive control process, analogous (though most probably not identical) to processes commonly believed to resolve conflicts in other interference tasks. Such an interpretation seems to be supported by neuroimaging studies showing that human brains react differently to targets than to lures, with the latter activating regions implicated in resolution of conflicts (e.g., Burgess et al., 2011; Chatham et al., 2011; Gray et al., 2003). In our view, the individual efficiency of some kind of executive control did influence the individual false alarm rates to highly distinguishable lures (while there was a lesser involvement of control in the rejection of poorly distinguished lures, because, as they were often taken as targets, they did not cause a high level of conflict). However, the presented experiments do not allow for identification of what the control process responsible for coping with conflicts related to lures is exactly, and examination of the characteristics of such a process surely needs a dedicated study (see Chatham et al., 2011; Ecker, Lewandowsky, Oberauer, & Chee, 2010; Szmalec, Verbruggen, Vandierendonck, & Kemps, 2011).

Here, only speculation on the role of control in reasoning can be drawn. If the attentional capacity allowing for information processing during reasoning is so limited, then, while dealing with difficult problems, people surely have to break them down into several smaller sub-problems and then build the final solution from the respective partial solutions (this is exactly how the influential LISA model of reasoning works; see Hummel & Holyoak, 2003). So, efficient control may be needed for the composition of final solutions. For example, while choosing appropriate sub-products (e.g., hypotheses, relations) one usually needs to avoid superficially matching mental representations that are dominant but not relevant to the problem (e.g., ones consistent with common sense but not following the premises). Instead, a problem solver may have to retrieve representations from memory that violate some well-learned schemata and semantic relations while still being goal relevant. The process of rejecting lures during performance on the n-back task could be somehow related to the more general control processes required for fluid reasoning, which allow for goal-directed selection of information. Due to better control processes of this kind, some people may be both more efficient problem solvers in the fluid reasoning tests and more accurate participants in WM tasks requiring the rejection of distractors. Such speculation seems to be supported by recent data indicating that proper selection and rejection of features of structures processed during reasoning is crucial for finding correct solutions (e.g., Cho et al., 2007; Chuderska, 2010).
Some Alternative Explanations of the Effect of n on the Correlation of Hit Rate and Reasoning

At least three alternative explanations of the data indicating the difference in correlation between Gf and hit rates on smaller versus larger n positions are possible. One potentially plausible explanation, already noted, refers to the more frequent use of the phonological loop by intelligent people, as the phonological loop is an important mechanism in some WM models (e.g., Baddeley & Hitch, 1974). However, in light of knowledge on the effects of rehearsal on WM performance, this hypothesis seems less likely than the one relating to the limits of the focus of attention and the effectiveness of control. Rehearsal, as a relatively passive mnemonic mechanism, blurs rather than increases the differences in working memory performance. As far as we know, no published results have shown that intelligent people use the phonological loop more often or more efficiently. On the contrary, it is often suggested (e.g., Cowan et al., 2007) that rehearsal in WM tasks decreases the strength of relation between WMC and intelligence. Moreover, in Experiments 3 and 4, sufficient time (2.8 s) for rehearsing the two recent stimuli was provided, but it did not influence the correlation between 2-back accuracy and reasoning compared to the remaining experiments.

Another explanation concerns the more efficient recollection processes of intelligent people.2 Recollection is believed to be a relatively slow mode of access to memory, which consists of the retrieval of all the associated information about an object (e.g., its specific features like color or size) as well as the contextual data associated with it (e.g., its position in the memory set). Recollection contrasts with the relatively fast, familiarity-based process of recognition, which yields only a general, unitary value indicating the confidence that the given object has or has not been coded in memory (Yonelinas, 2002). One possible explanation of the presented data states that due to the more efficient retrieval of contextual information about the precise n-back position of a stimulus (i.e., more efficient recollection), intelligent people accepted more targets and rejected more distractors than did less intelligent ones. No such difference would pertain to a familiarity-based process. However, this hypothesis does not need to contradict the hypothesis of limited attentional capacity, because recollection is often believed to only make use of the focus of attention, while familiarity processes probably operate on the whole WM, including its activated long-term components outside the focus (Oberauer, 2005). Assuming that reasoning is related to recollection, but recollection operates only on the focus of attention, which varies in capacity among individuals, would yield predictions analogous to those following the assumption of the limited and varied capacity of the focus underlying Gf. As our work was not aimed to precisely examine the processes occurring in the n-back task performance, but only to generally relate the structure of WM that is most commonly advocated in the literature, we only acknowledge the former, more specific explanation of our data and stay with a more general interpretation. An examination of exact mechanisms due to which n-back items are better processed in the focus of attention than outside of it requires a dedicated study.

The last alternative explanation that we considered regards possible differences in coping strategies in the n-back task. The use of either low-control or high-control strategies by participants in this task was suggested on the basis of both post-experimental questionnaires (Lovett, Daily, & Reder, 2000) and behavioral indices (Juvina & Taatgen, 2007). As less intelligent participants might have been aware of their worse cognitive functioning and could have perceived the n-back task as very difficult, they might have decided not to “invest” their attentional resources but, instead, to exploit some less demanding (but also less effective) processing strategy. Although we cannot fully reject this alternative, as no control over strategies was exerted in our studies, it should be noted that in Lovett et al.'s (2000) study, participants reporting a low-control strategy in the n-back task displayed worse performance (than high controllers) in cases of the largest memory load (i.e., the 3- and 4-back conditions) and not in cases of low memory load (i.e., the 1- and 2-back conditions). In Experiments 1–5, we found the opposite results in this regard. Strategy use and adaptation within WM tasks should surely be intensively studied (for an example, see Braver et al., 2007), but any contribution of processing strategy to the pattern of correlations found in our work seems unlikely.
Conclusion

The n-back is a relatively simple task, but processing in this test involves numerous cognitive processes and is very difficult to analyze (Ecker et al., 2010; Oberauer, 2002; Szmalec et al., 2011). However, on the basis of the complex pattern of correlations between reasoning and the n-back task performance observed in our experiments, we could identify the two factors, namely, the capacity of attentional maintenance of information, as well as the efficacy of control over distracting representations, which crucially contribute to the level of performance on this task. Most important, the individual differences in both these factors also contributed to fluid reasoning, and they did so in a relatively independent way. So, most probably, one may find people having both capacious WM and some problems with control over WM, as well as low-capacity efficient controllers. An interesting question for future research is whether deficits in one factor can be overcome with the other factor. For example, maybe some low-capacity people are able to carefully select information that enters their foci of attention and then to avoid wasting their scarce capacity on information unrelated to the current task, and vice versa, maybe the cognition of high-capacity individuals can suffer less from breakdowns of control.

Although the aim of this article did not consist of testing any theoretical model of WM, the complex pattern of observed correlations between Gf and our WM measures is consistent with the “standard model” of WM (O'Reilly, 2006), which assumes at least the bipartite structure of WM and the control processes operating upon it (e.g., Cowan, 1995; Engle & Kane, 2004; Unsworth & Engle, 2007; Oberauer et al., 2007). Each of these presumed elements of WM yielded a different relation to reasoning: from the strong one in the case of the focus of attention, through the moderate/weak one in the case of control processes, to the zero correlation in the case of activated LTM.F

[jotaro]

look pontus i agree with you,i believe that the guys who score high on iq and dont show working memory activity goes to show that those guys probably have the tasks of the test in the long term memory in other words their familiar with the task before hand ,and its like saying look he already knows the material so he must be a genius. but we all know material can be learned.

so iq test are useless.

but your last post is way too long man. :)

[Brandon Woodson]

"Quite frankly I am quite tired of the "Anti working memory movement" and all their logistical and almost religious explanations"

Why?

There is an obvious distinction to be made between Gf can be invariably and directly predicted by WM performance and Gf is strongly linked to WM performance in most cases.

The only way to be completely satisfied with a probabilistic approach to understanding this problem is if we consider that Gf isn't deterministic, which *probably* isn't the case; so it's natural that some would theorize about the inconsistency.

Also, the study doesn't say much that hasn't already been said - i.e., WM predicts *fairly* well in certain situations without giving any ameliorating explanation as to why there is inconsistency. Heck, there isn't even consensus among the researchers on what WM tasks are measuring. If it has anything to do with attentional control, as I've mentioned in the past, and WM task performance serves as a function of attentional control (instead of pure WMC), this would show a bias in testing procedure and explain the gap in theory - i.e., why there are high I.Q. persons with poor WM task performance on certain tasks who perform well on certain *other* WM-dependent tasks (like I.Q. tests) - especially when the type of WM being tasked comes into question.

Anyway, there are questions to ponder to which the cited studies hardly provide resolute answers.

-Brandon

P.S. I tried to post earlier, but my posts were twice automatically deleted.

[Closed]

It's important to remember that early working memory measures, such as Daneman and Carpenter (1980), were validated by their correlation with IQ-esque tasks (SAT, etc..).  While providing a useful basis for differentiating these early WM measures, it creates a kind of circularity when considering how WM measures relate to IQ, since the presence of that relation was (and often still is) a defining characteristic.

However, that's not saying anything for visual working memory measures (change detection tasks, etc..), or many other variants.  Studies that look at the effects of lesions to different brain areas on WM and IQ seem like another good avenue into the topic.

[Payman Saghafi]

Wait a minute!

I agree with Brandon.  My original post was not designed to resemble anything similar to an "anti working memory" manifesto. It is clear that working memory is an important construct; still, being an "important" construct does not make working memory the "primary" construct.

Pay

[Pontus Granström]

Very true, however, working memory is probably a better and purer way
of predicting academic and work potential since it's less influenced
by things like social class and other environmental issues.
I posted an article on this, it seems that WM is the new IQ. Since
it's so vital for actually getting things done.

On Thu, Jan 17, 2013 at 11:39 AM, M.M. <vividem...@aol.com> wrote:
> Of course some people have a high IQ and poor working memory. Since there is
> no such singular force as "G", it is no surprise why there is so much
> discrepancy among subtests.
> "G" is just a statistical probability of moderate strength. Everybody is
> certainly not around the same [er,performance level in separate tasks
> because each skill is the product of a different part of the brain.
>
> I remember there was a man who posted here (I don't know if he still does),
> but his IQ was high enough to join Mensa, though he actually scored in the
> high 70s in working memory! I remember him saying college level reading was
> very taxing for him due to this problem. Regardless, he use has excellent
> performances in other areas, qualifying him with a high IQ.
>
> Then there was another guy here was worried about his IQ being only 98. He
> scored below average in spatial and visual tasks, but I recall his verbal IQ
> being well above average, a bit over 120 if I remember correctly.
>
> If training tasks are tested to impact IQ in any way, they should first be
> tested in the actual subtest that the training is revolved around. Too many
> people here are worried about improving this theoretical "G" when they
> overlook gains in separate areas. If an improvement in made in ANY form of
> cognitive, it should be celebrated.

>
>
> On Thursday, July 5, 2012 3:37:55 PM UTC-4, Payman Saghafi wrote:
>>
>> We always hear about the correlation between IQ and working memory.
>> I'm wondering if there are exceptions to the rule in which a person
>> with a very high IQ has poor working memory. I have read a little
>> research about this, but would REALLY appreciate some additional
>> information.
>>
>> Pay
>

[M.M.]

I concur. 

If improvement in daily life is desired, their plan should encompass working memory, motivation, focus, and creativity. 

Don't get me wrong; overall IQ is certainly important and a worthy asset to have, but it's not everything.

Off Topic: If someone editing my posts? Upon re-reading my previous reply, I noticed there are extra characters I know I did not type (eg: "[er,")

[Michael]

Look at the location of "[" on the keyboard. You were starting to type 'performance'. Can't say much about the _'_ though, apart from the fact that its right next to 'm'; although, going by the position of the 'm' in 'performance', it seems like a false start, if indeed it was attempting to be in the race at that point.

[jotaro]

if thats the case as with pencil and pen, high iq and low wm

it just means iq is a false measure to begin with and you shouldnt realy on it typically.

wm is a more core ability then gf. also in that test of gkacademy it isnt exactly wm either because of guys might be familiar with those kinds of problems, taking all the challenge away.

[Richard Treadwell]

Hi Sonia, don't know if this thread is still active. Nevertheless, I just came across an article that stated working memory is more strongly correlated to academic performance than working memory. So, an initial comfort some parents encounter when receiving their child's IQ score can be quickly offset by other factors. However, working memory can be improved, and recent studies have shown transfer effects to other areas, many of which are tested with an IQ exam. I'll post some links if this post gets a reply (I'll know the thread is still active.)

Richard

On Saturday, March 16, 2013 8:04:38 AM UTC-7, Sonia Heidrich wrote:

I'm doing research lately on working memory online and I came upon your question. My son has a high IQ and very low working Memory. Do you have any info about that? He is very good in Math and reads fairly well but gets low grades in class because he leaves work unfinished and teachers don't know how to handle this. He is 10 years old.

Sonia Heidrich

On Sunday, July 8, 2012 10:19:18 AM UTC-4, Green wrote:

You might search the literature on TBI to find examples of such people.

[Richard Treadwell]

*working memory is more strongly correlated to academic performance than IQ. (Obviously have to work on my working memory, ha.)

[Pontus Granström]

Executive functioning in general is a better predictor of most things
that we would associate with intelligence.

> --
> You received this message because you are subscribed to the Google Groups
> "Dual N-Back, Brain Training & Intelligence" group.

> To unsubscribe from this group and stop receiving emails from it, send an
> email to brain-trainin...@googlegroups.com.

> To post to this group, send email to brain-t...@googlegroups.com.

> Visit this group at http://groups.google.com/group/brain-training?hl=en.

> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>

[Oppo Thumbs]

I have a high IQ and a very poor memory, particularly visual memory.  I can repeat 8 numbers forward or I could years ago but at best 4 digits backward. It must be extremely rare wretched condition to have. A neurologist told me I'd be better off with a below IQ and average memory.  

[Oppo Thumbs]

You are very lucky. I have a verbal IQ of 125 but a 75 memory IQ. You will never know how lucky you are.

On Monday, July 9, 2012 11:41:22 PM UTC-4, krs wrote:

On Thursday, July 5, 2012 2:37:55 PM UTC-5, Payman Saghafi wrote:

We always hear about the correlation between IQ and working memory.
I'm wondering if there are exceptions to the rule in which a person
with a very high IQ has poor working memory.  I have read a little
research about this, but would REALLY appreciate some additional
information.

I think I'm the guy you're looking for.

On most of the subtests in the SB battery my IQ ranges from the upper-120s to
mid-130s. On the subtests for memory my IQ ranges from the upper-70s to the
mid-90s. People like me are considered learning disabled for this reason.

In comparison, when I took the MAT for grad school my score gave me an IQ-equivalent
in the upper 160s. Getting in to grad school was easy, but getting out was hard.

[Vanessa Perry]

Is it far too late for me to respond to the question, or are you willing to hear my say?

On Thursday, July 5, 2012 at 8:37:55 PM UTC+1, Payman Saghafi wrote:

We always hear about the correlation between IQ and working memory.
I'm wondering if there are exceptions to the rule in which a person
with a very high IQ has poor working memory.  I have read a little
research about this, but would REALLY appreciate some additional
information.

Pay

[Brain Train]

@Vanessa

It's an interesting question. 

I am curious to know if you have something to add.

--

You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.
Visit this group at https://groups.google.com/group/brain-training.

For more options, visit https://groups.google.com/d/optout.

[Vanessa Perry]

The condition of ADD is bizarre. Persons with ADD can only focus their attention if they feel passionate about what they are learning, but seriously struggle when they have no passion. What’s going on here? What’s with this bias attention of theirs? My younger sister can’t manage to complete basic tasks like doing house chores. The whole family thinks she’s lazy and that she doesn’t care enough to help with the chores. But she’ll spend several hours straight in her bedroom writing these stories, and tells me to go away the moment I come into her room. How can she spend so much time writing but not be willing or able to do house chores? Sorry but it doesn’t add up. The ‘central executive’ is a myth. My sister undermines the entire thing…oh she can focus her attention when she feels like it. I want answers. I want to know what’s going on with my sister and with this condition of ADD.

The ‘central executive’ cannot possibly exist and those with ADD prove that it doesn’t exist. The brain must have two different forms of attention control: 1) for following orders, such as being give chores to do (where my sister seriously falls short), and 2) for working on something personal, such as writing stories (where my sister can spend up several hours straight writing them). Now this would not be possible if there really was a ‘central executive’. It obviously doesn’t exist. On the basis of this finding, it is difficult if not impossible to accept the idea of a central executive as defined by the working memory model. Instead, it appears that there is no single controlling resource, and that different executive tasks involve non-overlapping control processes. Working memory only accounts for attention control when it comes to following orders and carrying them out, there must be another form of attention control that is used when you are tasking while doing as you please. My conclusion is supported by those with ADD. They don’t lack the ability to pay attention all together, they lack the ability to pay attention when they are being given orders and have to carry them out. If someone tells them to arrange a bookshelf into alphabetical order, they probably won’t do it because the attention is not there, but if they desire to arrange the bookshelf into alphabetical order they will do it because now the attention is there. The desire to do something has a different form of attention control, and it’s not working memory. It’s something else and I’m hoping that maybe someone can give me an answer??

On Thursday, January 18, 2018 at 1:40:24 PM UTC, ADDer wrote:

@Vanessa

It's an interesting question. 

I am curious to know if you have something to add.

On Tue, Jan 16, 2018 at 3:03 AM, Vanessa Perry <nintype...@gmail.com> wrote:

Is it far too late for me to respond to the question, or are you willing to hear my say?

On Thursday, July 5, 2012 at 8:37:55 PM UTC+1, Payman Saghafi wrote:

We always hear about the correlation between IQ and working memory.
I'm wondering if there are exceptions to the rule in which a person
with a very high IQ has poor working memory.  I have read a little
research about this, but would REALLY appreciate some additional
information.

Pay

--
You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-trainin...@googlegroups.com.
To post to this group, send email to brain-t...@googlegroups.com.

[Pontus Granström]

I recommend reading about PFIT. I can send you a free ebook.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.

[Brain Train]

I guess voluntary control in ADD people is poor. Their behaviour is desire driven. I remember an ADD expert saying that ADD is not 'attention deficit disorder', it is 'Attention control deficit disorder'.

Challenge with ADD people is learning things which are not straight forward. Understanding new complex thing requires bigger working memory.

Paying attention to a movie or an interesting 'you tube' video may not be the real test of attention. Most of us can do that for hours together. It is not difficult to pay attention when we are curious about something, and it is easy to understand. Example- a child listening story from his grand-mother.

The challenge is to pay attention when there is no immediate reward and it is not straight forward either. 

For example, the research papers (like the ones this community shares here), they are lengthy and often not very clear or intuitive through out their narration. You have to wait while you read. you have to understand and absorb what the author is saying without knowing what will come out. It wouldn't be making sense as you read it and picture will be clear only towards the end. This is the kind of scenario, i guess, someone with ADD would find challenging. 

Because he can't easily control attention; he can't casually say to himself 'okay, this stuff is very boring but let us go thru it and understand what this study says'. They need a stronger reason (or higher reward) to go through boring stuff as compared to normal people who do not have ADD.

And it doesn't end here. If driven by high reward, they try to go through that tedious research paper.. they would have to spend longer time and greater effort. they may have to read it repeatedly to be able to understand it and to make sense out of it.

BT

[Vanessa Perry]

"Understanding new complex thing requires bigger working memory."

No, it doesn't. "Complexity" is associated with logic, pattern recognition, abstract thought and problem solving. Working memory is associated with simple rote tasking which is done in school. Ironically, many of those with ADD struggle to learn simple things because of the working memory deficit. However, like my sister, many of those with ADD can learn more complex things such as structuring stories. This is creativity. My sister definitely has good problem solving skills, I'll give her that. Don't you see what I am saying? My sister contradicts and undermines the idea that working memory determines the ability to learn when, in reality, it does no such thing. Instead, problem solving determines the ability to learn. The true measure of intellectual potential is Raven's Matrices. This is logic, pattern recognition, abstract thought and problem solving capacity. Seriously, if you want to find out someones real potential, just give them a Raven's Matrices test. That would be getting "straight to the point".

[William Zeller]

You don't understand what working memory is.  Please do some research before making another duplicative post.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.

[Vanessa Perry]

No, you fool. I understand perfectly well that working memory is equal to rote learning, the thing you do in school! Critical thinking is different and relies on pattern recognition. Are you all too slow to comprehend what I am saying? The answer is probably yes. Don't you understand? I am criticizing the working memory model because (in my eyes) it doesn't quite add up. If you don't agree then you must also criticize my responses and tell me exactly why you don't agree. That's how it works. Do you understand? 

[TranquiLogic]

What I've gathered about ADD is that it is an executive functioning problem involving the frontal lobes. The reason people with ADD can pay attention to things they are interested in is because excitement/passion activates their frontal lobe. They need this stimulation for their working memory to work better. As for ADD creativity, I think i read somewhere that frontal-temporal dementia can bring out a lot of creative ideas. Great ideas can come about sometimes when the mind is wandering. ADD and hypomania/mania are probably similar to the fronto-temporal dementia in this regard.

 

As for the Raven's Progressive Matrices test, it seems to me that you need to have a large working memory to be able to figure out the next pattern. If you do well on that test but say you have a poor working memory, maybe you just have a poor auditory working memory but superior visual memory. This test also has a lot to do with the functioning of your right hippocampus. I have a pretty good working memory but I don't do so well on that test. Maybe it is because my working memory strength is mainly auditory. I think it is just because I've been sleep deprived my whole life, went without sleep for two years recently, and am recovering from it. I know my hippocampus has got to be in terrible shape. 

Instead of aiming for just the common idea of intelligence. I think it is important to try to improve the health of the whole brain. The ravens test may be considered the most important test (it may be for fluid intelligence) but what about the other areas of the brain like the temporal lobes (emotion-long term memory visually and semantic), the cerebellum (balance, coordination, and probably language socialization because of its connection to autism), the parietal lobe (visualization, mental rotation), the right temporal parietal junction (cognitive empathy),bilateral anterior+insula anterior cingulate cortex+ inferior frontal gyrus (affective empathy), etc.   I wonder if there is anyone else here who is obsessed with improving the whole brain too. I heard of the room changing n back game that is supposed to improve episodic memory (hippocampus) as well as working memory. Has anyone tried this and have they noticed a difference in their long term memory? Also has anyone here tried anything that made them significantly better at seeing things in their mind's eye?? 

[William Zeller]

You said: "I understand perfectly well that working memory is equal to rote learning, the thing you do in school! "

Huh??

Working memory is not "rote learning."  That's long-term memory, which is a totally different thing and implicates different brain structures (a primer on the differences between long term, short term, and working memory here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2657600/).  

Fluid intelligence -- which Raven's is the gold standard test for -- is so highly correlated with working memory that some have argued they are the same thing.  (Here's the first study I found in a quick google search discussing the relationship among fluid intelligence, WM, and Raven's: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3968765/).

You've made the same shrill and inaccurate post arguing with nobody in particular about 5 times now and I'd prefer you stopped doing so.  

Thanks!

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.

[Vanessa Perry]

Bullshit. Raven's Matrices is separate from working memory. There are examples of people who score high on Raven's Matrices while having a low working memory. Similarly, there are examples of people with exceptional working memory despite having an average score on Raven's Matrices. This is called contradiction. You can't continue to believe that Raven's Matrices and working memory are connected with these kind of contrary examples. They are only found to be correlated which of course, as always, is not the same as being the "causation". The contrary examples prove that they are not directly connected to each other. Working memory only determines rote learning. Working memory simply has nothing to do with Raven's Matrices.

[William Zeller]

You are not making sense.  I'm not sure if you're trolling or actually think this way, but I'd urge you to read something and quiet down until you know what you're talking about on the off chance you care about your reputation.  Either way I'm finished talking with you.  

Good luck.

--

[Vanessa Perry]

I'm presenting a new idea and you can't handle how conflicting it is with the current idea. Have fun with your cognitive dissonance. Bye, bye, ignorant fool!

To unsubscribe from this group and stop receiving emails from it, send an email to brain-trainin...@googlegroups.com.
To post to this group, send email to brain-t...@googlegroups.com.

[TranquiLogic]

Here is a simple example of working memory. Someone gives you a list of grocery items to get: cheese, milk, water, kale, lettuce, pork, and beef. You could remember it in the order that it was told. That is more of a rote way of doing it. Or you could remember it by categories so it is much easier: vegetables (kale, lettuce), meat (pork, beef), water, dairy (cheese, milk). You can also remember the categories in terms of the layout of the store. This is how I would try to remember it. In order to remember that, you have to maintain the items in your head and at the same time search your long term memory for their appropriate classification. You may be in a noisy environment and be distracted. If your working memory is strong, you won't succumb to the interference. You could call this simple rote memory but it is more than that. You are organizing the items in your head under broader categories. You are also applying logic to the list by remembering the items in the categories to get in terms of the layout of the store. 

This is one of the everyday uses of working memory that you may call just simple rote memory. When it comes to the brain, it seems like all these type of "simple processes" must work efficiently before you can do more complex things. Problem solving doesn't even begin to take place before you can use logic. Logic doesn't work if you can't maintain thoughts in your head. You don't instantly know the logical way to do something. I used to think the smartest people just suddenly knew what to do. I now know after my brain healed, that everything is a struggle even if it is a very small struggle. 

Vanessa, if you are getting superior results in the Ravens Progressive Test, maybe it is because visual spatial tests get you excited. Maybe it is because you have high visual spatial working memory but poor auditory working memory. Or maybe you feel like you aren't using your working memory but you are somehow subconsciously using it. The answer to the problem comes to you without your awareness of how you worked it out. I would find this hard to believe but it might be possible. Or maybe you just have a high visual working memory and have worked out similar kind of problems so much that you are so efficient at it, it seems like the answer comes out of nowhere. Either way I would suggest that you try dual n back with interference control on normal speed. When you get good at it, then try crab n back with interference on normal speed. After you master normal speed, keep doing crab interference on increased speed. Once you master a fast pace, go on to the next crab n back level with interference on normal speed and continue this pattern. Keep doing this and give it time. I think it will help your working memory. This is important for so much more than just problem solving. It should help with impulse control, emotional stability, and giving you a calm state of mind. I would recommend you try hyperbaric chamber therapy because that did more for me than anything else. Either way, dual n back has been shown to help people.

[Vanessa Perry]

Not all people rely on working memory to learn, here’s why:

https://wrongplanet.net/forums/viewtopic.php?t=173095

http://www.macleans.ca/education/just-the-facts-heres-why-rote-learning-is-wrong/

This is called counter evidence. This is the contrary.

Apparently, those who think that everyone relies on working memory to learn are deeply confused.

[Pontus Granström]

Attention control is the gate through which all information must pass. It has nothing to do with creating hash maps with facts. In order to comprehend it needs to stay in the working memory long enough so you can comprehend it then store the comprehension in long term memory. Working memory is a bottleneck and a severe one as well.

--

You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.

[Vanessa Perry]

LOL, LOL AND LOL

Look, infantile moron, just fuck off and don't waste my time.

Intelligence = recognizing contradiction and refuting old ideas

Lack of intelligence = holding on to false beliefs for as long as possible

To unsubscribe from this group and stop receiving emails from it, send an email to brain-trainin...@googlegroups.com.
To post to this group, send email to brain-t...@googlegroups.com.

[Vanessa Perry]

King Of The Stars is completely disillusioned by the working memory model. He/she has been so for years it would seem. They're a broken record. Their response is always an emotional one and never a logical one. Intelligence will never be defined by attention spans which, of course, alternate all the time. How absurd. Anxiety alone will wreck your attention. Intelligence is defined by the capacity for abstract thought. Your argument makes perfect sense by the way ;)

On Tuesday, November 20, 2012 at 8:43:59 PM UTC, Brandon Woodson wrote:

"Quite frankly I am quite tired of the "Anti working memory movement" and all their logistical and almost religious explanations"

Why?

There is an obvious distinction to be made between Gf can be invariably and directly predicted by WM performance and Gf is strongly linked to WM performance in most cases.

The only way to be completely satisfied with a probabilistic approach to understanding this problem is if we consider that Gf isn't deterministic, which *probably* isn't the case; so it's natural that some would theorize about the inconsistency.

Also, the study doesn't say much that hasn't already been said - i.e., WM predicts *fairly* well in certain situations without giving any ameliorating explanation as to why there is inconsistency. Heck, there isn't even consensus among the researchers on what WM tasks are measuring. If it has anything to do with attentional control, as I've mentioned in the past, and WM task performance serves as a function of attentional control (instead of pure WMC), this would show a bias in testing procedure and explain the gap in theory - i.e., why there are high I.Q. persons with poor WM task performance on certain tasks who perform well on certain *other* WM-dependent tasks (like I.Q. tests) - especially when the type of WM being tasked comes into question.

Anyway, there are questions to ponder to which the cited studies hardly provide resolute answers.

-Brandon

P.S. I tried to post earlier, but my posts were twice automatically deleted.

[Brandon Woodson]

Vanessa, not trying to be oppositional here; I'm genuinely curious. For you, what is attention? Does it play any role in intelligence whatsoever? If so, what?

--Brandon

--

[Vanessa Perry]

Working memory is NOT the 'primary construct' except for rote learning in school. It is the primary construct in that regard. Persons with low working memory really do struggle to learn by rote, especially times table and calculation. They can't hold the numbers in their heads long enough, just like how they can't memorize much factual information. Indeed, their memories won't hold it or keep it retained. My sister has ADD and she failed to learn her times table.

On Sunday, November 25, 2012 at 7:38:09 PM UTC, Payman Saghafi wrote:

Wait a minute!

I agree with Brandon.  My original post was not designed to resemble anything similar to an "anti working memory" manifesto. It is clear that working memory is an important construct; still, being an "important" construct does not make working memory the "primary" construct.

Pay

[Brandon Woodson]

Sure, I understand that. I personally know of examples like you described. But what is attention to you? Does it relate to intelligence? If so, how?

--Brandon

--

[Vanessa Perry]

There are two forms of attention control 1) for following orders and carrying them out, as defined by working memory. The ability to learn by rote, particularly in school, and 2) for personal desire. When a person has the desire to learn how to do something, say, structuring stories, they use a different form of attention control. I just don't know which part of the brain it is but I'm supposing it is pattern recognition and logic. This explains why persons with ADD can't learn well in the classroom but can somehow learn well on their own. My sister can spend hours upon hours working on her stories, but can't spend 15 minutes filling out a form. She just loses focus. Don't you see how bizarre this is? Her attention span alternates depending upon the situation of learning or working. ADD is a bizarre condition.

[Brandon Woodson]

Yes, there appear to be different mechanisms that control attention. What precisely is being controlled? Or, in other words, what is attention itself? Where does that fit into the picture of intelligence?

[Vanessa Perry]

I'm not entirely sure what you're asking me. Attention is focusing while intelligence is thinking, of course they work together. Look at it this way, humans comprehend in patterns and make sense of their ideas by applying logic to them. This is a process whereby you look for errors and contradictions and then carefully weed them out, so that you're left with a consistent, logical pattern of information. The more errors and contradictions there are in an concept, the less it makes sense. Making sense of things then always involves this abstract mental process of looking for errors and contradictions. Comprehension is determined by pattern recognition. Intellectual potential can be measured directly by giving a Raven's Matrices test. 

A working memory deficit is a spatial memory deficit, a small, shrunken or undersized hippocampus. A working memory deficit is NOT a complete deficit in attention, but that is what is being supposed by idiots. The working memory model is based on the idea that there is "central executive" that controls all major brains functions, which is ultimately wrong. If this were case then those with ADD would have a complete deficit in attention. They would not be able to function at all without taking stimulant drugs every single day. This is clearly NOT the case. If there is no "central executive" then the working memory model is a false construct.

I think it is more likely that short-term memory works together with pattern recognition. Short-term memory activates the information and pattern recognition does the comprehension, the problem solving. This is is the ability to learn information conceptually. Whereas working memory is the ability to learn by rote, to memorize factual information, such as, 4 multiplied by 5 equals 20. There is no logic here, no understanding. Its just memorized. Owls are birds of prey, their heads can turn almost 360 degrees, and they eat field mice. Facts like these are also memorized using working memory. None of it requires you think very hard. That's my theory anyway.

So, what I've basically argued is: working memory is crucial for rote learning, and short-term memory is crucial for problem solving along with pattern recognition. Working memory is NOT crucial for the whole brain. There is NO "central executive".

[robert chalean]

Your sister can convert the numbers to images to better retain them. http://memory-sports.com/blog/memory-techniques/the-major-system/ 

[robert chalean]

Your sister can use this app. She have to follow the balls. http://competicionmental.appspot.com/router?page=nt3d2

[Shoh]

Hello, Vanessa, you are very much spot on, except this part, I don't agree with:

"So, what I've basically argued is: working memory is crucial for rote learning, and short-term memory is crucial for problem solving along with pattern recognition. Working memory is NOT crucial for the whole brain. There is NO "central executive". "

The reason is very simple, because pattern recognition, rote learning, mental calculation all are considered forms of memory, all of them in fact are memories. Even, in chemical reactions, they are almost essentially the same, whether you engage in abstract thinking, rote learning, or calculating two digit numbers in head - the fact is in these all cases, a brain receives influx of bits of memories. 

Based on above, what scientific community thinks is if a person has good central executive function, and possibly, he/she will have similarly good performance in all problem solving areas.(But no way, working memory makes up for pattern recognition, if someone has excellent memory, but poor pattern recognition library, then it is no avail)_ In other words, if someone is good at managing his memories, then he/she is also good at managing memories related to pattern recognition, abstract thinking and rote learning. As humans do think in memories, and everything brain processes is memory, then it is okay to reason this ability to manage memories efficiently should be transferable to all other cognitive proccesses, which all are the same things, but different contents - in short, memories.  

Regarding to ADD, why ADD person can do things discriminatorily efficiently in certain areas, while struggling in other areas, this may to do with 

long term memory - that is by doing things over and over again, involuntarily loading all information related to this much repeated processes into long term memory, a person with ADD, can carry out the specific tasks with no problem. Another possible reason, human brain works in analogies, in habits, so bias or interest to some areas may be from personality of a person - he or she formed certain long term memories that make her or him good when person is engaged in those activities.

[Hi]

A simplified dichotomy to remember is that comparatively, memory is representational and fluid intelligence is relational. 

Ties in well with very intelligent people with ADD.

Any output of intelligent activity is due to relations being able to be represented in the mind long enough that they could be executed. Any output of activities exercised by memory is due to passing the most minimum of thresholds regarding what constitutes intelligent activity but not necessarily any greater than this let alone in the higher ends. 

The relational reasoning n-back game trains both components.

[uyghui ghbbjb]

What relational reasoning game is this?

[Gore Lando]

I'm not an IQ fetishist but I scored 160 on the old Stanford-Binet Intelligence test when I was a kid, the maximum for that version of the test. And I have severe inattentive AD(H)D with assumedly poor working memory as a result of that condition.

My life is sheer suffering by the way. I wouldn't recommend having a kid if you think they'll have ADHD and be clever, because it feels even worse to fail when you're clever.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.

[Brain Train]

@Gore Lando

ADHD related issues can be controlled by medicines and other interventions, to some extent. I guess, it can make life manageable, liveable.

It's upto the patient (and people concerned- family, friends, etc) to recognise the need and opt for the treatment.

I have some experience of debating with some of these people who I believe may be having undiagnosed ADD. These are toughest people to convince... they will not even accept that they have some problems in their lives due to their unusual behaviour. If they take some initial stand on any debate, they usually get stuck with that even if facts are presented which clearly contradict their conviction.

They find it hard to accept that they were wrong! They are very creative and keep coming back with one argument or other. Also, they tend to jump from one point to other point of debate when they see they are approaching dead end on their line of argument!

So, ADHD symptoms can be controlled if they really wish... but bigger issue is that they usually don't recognise it, and hate taking any medicine for it.

BTW, ADHD people can shine if they pick the job/activity which is aligned with their area of interest. 

They need to be much more careful on what they choose to do as compared to other people (because either they will shine or they will fail... no average performance usually!).

-BT

[robert chalean]

Today a i read on Quora that if fou have ADD an do a IQ test and you do not take medication may be you can have a score of 130 and with medication you can have a score of 160.

[Brain Train]

Interesting info.

BTW, it looks like ADD people are unusually creative and IQ test doesn't test this unique strength!

So, IQ score may not be reflecting the true potential of ADD people.

So, people with ADD shouldn't dishearten even if IQ score is lower than their expectation. Their real strength lies beyond IQ... and that is their ability to find new and unconventional solutions through their out-of-box thinking capabilities!

On Thu, Feb 22, 2018 at 7:58 AM, robert chalean <robert...@gmail.com> wrote:

Today a i read on Quora that if fou have ADD an do a IQ test and you do not take medication may be you can have a score of 130 and with medication you can have a score of 180.

To unsubscribe from this group and stop receiving emails from it, send an email to brain-training+unsubscribe@googlegroups.com.
To post to this group, send email to brain-training@googlegroups.com.

[Individual whom enjoys watching Rick and Morty]

I'm the person you're looking for, it's completely phenomenal to me that I wasn't able to find anybody online with the same problem I face, I am a perfect example of that problem, overall my IQ is about 130, almost Mensa level, my visual/spatial IQ is 140 and my verbal is about 100, whereas my working memory is complete utter garbage and makes me feel extremely insecure and doubtful of my overall cognitive abilities, learning is an immense difficulty, when studying, I have to repeat something about 30 times, no exaggeration, whereas one of my close friends sitting next to me in class has an amazing working memory, can read something once/twice and be able to recall 90% of the information with ease, he of course has straight A's, I am completely impaired due to the fact that I am cognitively cursed, I am intelligent in one aspect and not in another, my existential intelligence is also very high, never did an exact test, although I am aware of the fact it's high due to my dread for philosophy and existential thought. If there is anybody on this forum like me, please reach out, I'd love to hear your experiences, and if anybody's got some questions, just ask me.