------------------------------------------------------------------------------
Title: (2009) Storage and non-storage components of working memory
predicting reasoning: A simultaneous examination of a wide range of
ability factors
Intelligence, Volume 37, Issue 4, July-August 2009, Pages 347-364
Abstract
The current study examined basic cognitive abilities that are related
to or included in the concept of working memory (WM): different WM
components, three executive functions, simple short-term storage
(STM), and sustained attention. Tasks were selected from well-
established models and balanced in terms of content. The predictive
power of storage and non-storage components of WM was examined. The
same analyses were conducted with mental speed as an additional
predictor of reasoning. Finally, we investigated whether the
identified predictors of reasoning remain relevant when a general
factor is considered in the analysis. The analyses revealed that short-
term memory accounted for the relationship between complex span
measures of WM and reasoning but not for the relationship between
coordination and reasoning. These findings remained stable in the
context of a mental speed and a general factor. The mental speed
factor did not add an incremental contribution to the prediction of
reasoning above and beyond short-term memory and coordination. The
general factor was mainly built by mental speed tasks and acted as an
additional predictor of reasoning besides STM and coordination.
----------------Thus, we concluded that reasoning ability can be
explained by coordination, STM, and mental
speed.------------------------
On Sep 22, 11:40 pm, genvirO <carsthatdr...@hotmail.com> wrote:
> Here's the discussion
> ------------------------------
>
> Discussion
>
> The present study revealed an important new result: Controlling
> crystallized intelligence did not have an impact on the explained
> variance betweenWMand reasoning. Moreover, the study replicated that
> WM and sustained attention together account for about 83% of reasoning
> variance. Also consistent with previous findings, we confirmed that
> storage in the context of processing and coordination were significant
> predictors of reasoning. As hypothesized, we could show that
> coordination and sustained attention were highly correlated. Finally,
> the correlations between all factors can be explained by a speed and g
> factor. All tasks showed significant loadings on the speed factor,
> while only storage in the context of processing, coordination, and
> intelligence tasks loaded significantly on g. If we control for g, we
> can explain the remaining correlations by speed. Supervision predicted
> speed and storage in the context of processing. The first goal of the
> present study was to confirm that the relationship between WM and
> reasoning does not decrease if we control for crystallized
> intelligence.As pointed out by Oberauer et al. (2005), there is a
> mismatch between WM and intelligence constructs. It was assumed that
> intelligence relies on conceptual structures while WM does not.
> As a consequence, the relationship between WM and reasoning should not
> decrease if crystallized intelligence is controlled. This could be
> shown within the present study. The explained variance of WM and
> sustained attention factors predicting reasoning was 83% and thus
> lower than that
> obtained by Buehner et al. (2005). But it did not change when
> crystallized intelligence was controlled for. In addition,
> the central findings by Buehner et al. could be replicated: The
> factors coordination and storage in the context of processing were
> significant predictors of reasoning. However, the importance of the
> predictors was slightly reversed: Coordination seemed to be a more
> important predictor of reasoning than storage in the context of
> processing. This might be the result of sample fluctuations. All in
> all, two factors of WM serve as significant predictors of reasoning:
> coordination and storage in the context of processing. Since construct
> reliability of WM turned out to be low, we applied an additional path
> analysis to predict reasoning using aggregated z-scores.
> Unfortunately, aggregates in this special case have the undesirable
> property that content variance is mixed up with variance from
> functional facets (positively correlated error variance). This leads
> to systematically increased variances of the aggregates. Despite this,
> and despite a higher reliability of aggregates, the variance between
> WM and reasoning was only about 50% (see also Buehner et al., 2005;
> Wittmann & Suess, 1999). This is exactly the amount assumed by Kane et
> al. (2005). However, they probably based their assumptions on
> correlations between latent factors and did not consider coordination
> in their calculation. In contrast to this, Ackermann et al. (2005)
> claimed that the common variance between intelligence and WM is much
> lower than suggested by Kane et al.. This cannot be confirmed within
> the present study: using a carefully selected set of tasks for two
> central WM components (coordination and storage in the context of
> processing) led to more common variance between WM factors and
> reasoning than reported by Ackermann et al.. Taking into account that
> the sample in the present study was homogeneous and, thus, restricted
> in range, the common variance between WM and reasoning might even be
> underestimated. All in all, it is reasonable to assume that
> the explained variance of reasoning predicted by WM obtained in
> multiple regression analyses probably represents a lower bound. One
> might argue that the results of confirmatory factor analyses in the
> present study suggest that WM and reasoning were almost identical when
> controlling for measurement error: WM and reasoning shared about 83%
> of variance. This result is clearly in line with findings obtained by
> Buehner et al. (2005), Colom et al. (2004), and Colom, Abad, Rebollo,
> and Shih (2005). Again, considering that our sample was restricted in
> range, one might suppose that both constructs are indeed identical or
> isomorphic constructs. But we do not believe that: As pointed out, SEM
> results lead to an overestimation of the common variance between
> reasoning and WM when the construct reliability (of the factors) is
> low. This might have occurred since the communality estimates in SEM
> were dramatically lower than the reliability estimates calculated by
> Cronbach's α.
> Things get even more complicated since Cronbach's α represents an
> upward-biased estimate of reliability when correlated errors occur as
> it is the case in our models. If measures are tau-congeneric,
> Cronbach's α is biased downward. This is also the case in the present
> study. The two biases might cancel each other. Regarding construct
> validity, the reliability estimates were considerably higher than
> communalities, indicating that the true reliability might be
> considerably higher than communality estimates in SEM but lower than
> Cronbach's α. Thus, SEM results can lead
> to an overestimation of correlations between latent variables. This
> was the case in the present study and probably holds true for several
> other studies as well (see Buehner et al., 2005). Moreover, if a
> correlated trait correlated uniqueness (CTCU) model is applied, the
> trait factor loadings are likely to be larger than for correlated
> trait correlated method models (CTCM; Lance, Noble & Scullen, 2002).
> Since we used CTCU models to ensure model identification, it is
> reasonable to assume that loadings are biased upward. Thus, the
> explained variance of reasoning might be overestimated (see also
> Lance, Lambert, Gewin, Lievens, & Conway, 2004). We believe that, for
> these reasons, WM and reasoning are not identical. All in all, we
> believe that Oberauer et al. (2005) made a reasonable estimate of how
> strongly WM and reasoning overlap, namely, about 70%. Another goal was
> to clarify the relationship between coordination and sustained
> attention. Through SEM and path analysis it could be confirmed that
> sustained attention and coordination are highly correlated. The
> correlation might even be larger for the following reasons: First of
> all, in our study the coordination tasks differ in several ways from
> the applied sustained attention tests. Three sources of method
> variance might have reduced the correlation between coordination and
> sustained attention: The d2 was administered as a paper-and-pencil
> test, the factor sustained attention contained only figural content,
> and the factor coordination consisted of four tasks, while sustained
> attention of only two (lack of symmetry). Consequently, future studies
> applying balanced sets of coordination and sustained attention tasks
> should reveal an even higher correlation. If that were the case, how
> should we interpret this similarity of concepts? Do sustained
> attention tests and coordination tasks simply assess an integration
> function? Or are coordination and sustained attention linked by speed?
> The present study reveals that the variance shared by coordination and
> sustained attention results from speed and not from an integration
> function. If we control for speed, only a small correlated error
> variance between both factors remained which failed to reach
> significance. This might serve as explanation for results found by
> Schweizer and Moosbrugger (2004), whose study revealed an incremental
> validity of the FACT in predicting reasoning above and beyond storage
> in the context of processing. Our results suggest that this finding
> might have occurred because of speed. We also assumed that speed might
> be responsible for the correlation between storage in the context of
> processing and coordination. This was confirmed within the present
> study. The variance not attributed to g can be attributed to speed.
> All other cognitive task aggregates also have significant loadings on
> speed. Speed in turn can be predicted by supervision, which might
> enhance the application of speed. As mentioned in the Introduction,
> the application of speed improves the performance in speeded cognitive
> tasks. This also holds true for storage in the context of processing.
> However, there also exists a direct path from supervision to storage
> in the context of processing. This is in line with Kane and colleagues
> and can be explained with attentional control. Unfortunately, the WM
> model applied in the present study does not include inhibition. On the
> basis of the proposals by Kane et al. (2005), it is reasonable to
> implement inhibition into this model. Concerning the last model
> presented, it is important to remember that g is a very illusive
> construct and is measured differently in every study. If we disregard
> this aspect, we might conclude that WM is g plus speed. However, they
> only explain between 50% and 60% of WM variance. Looking at the
> reliability ofWMaggregates, there is either much specificity left or
> much more to be explained. Colom et al. (2005) showed that one source
> of variance neglected in this study is short-term memory (STM).
>
> Thus, WM might be decomposed in variance from reasoning or g, variance
> from speed, variance from STM, content variance, and functional
> variance (the latter two according to Oberauer et al., 2003). This
> directs the way of future research: Only a model that includes all
> possible sources of variance could enhance our understanding of
> cognitive abilities and their interplay.
>
> On Sep 22, 11:23 pm, genvirO <carsthatdr...@hotmail.com> wrote:
>
>
>
>
>
>
>
> > This is interesting stuff! If anyone wants any of the articles just
> > let me know.
>
> ...
>
> read more »
--
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-training@googlegroups.com.
To unsubscribe from this group, send email to brain-training+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/brain-training?hl=en.
No comments:
Post a Comment