The aspects of intelligence that most interest us relate to the Cattel- HornCarrol (Carroll, 1993 ; McGrew, 2005 ) theory, and specifically to the distinction between fluid and crystalized intelligence. This distinction dates back to the work of Donald Hebb (see Brown, 2016 ), who distinguished between intelligence A (fluid) and B (crystallized; Hebb, 1942 ). Beginning Zach Shipstead & Randa l l with his work on the relation between brain damage and specific intellectual impairments, it has come to be known that damage to the prefrontal cortex can impair fluid intelligence (Duncan, Burgess, & Emslie, 1995 ), the ability to reason with novel information. Conversely, crystalized intelligence, a person’s accumulated knowledge base that can be put to use in routine problem- solving situations, is unaffected by such damage.
A long- standing issue in fluid intelligence research is the need to define its nature at a deeper level than “novel problem- solving ability.” In other words: what are the cognitive mechanisms that account for novel reasoning? In recent years, a chief explanation of differences people show in fluid intelligence has been through appeal to its correlation with working memory capacity. Working memory is a cognitive system that allows people to maintain and manipulate information in the service of ongoing cognition. Capacity refers to individual differences in the efficacy with which this system functions.
Individual differences in working memory capacity (as it is typically defined) and individual differences in fluid intelligence are strongly predictive of one another. It is well established that these factors share 50– 80% of their variance at the latent level (Chuderski, 2013 ; Kane et al., 2005 ; Oberauer et al., 2005 ). Thus, research has been progressing under the assumption that when the working memory system is understood, this knowledge can be applied to a large portion of fluid intelligence.
As the story goes, working memory has a causal influence on reasoning ability. Stable maintenance in working memory allows people to focus on relevant information (Engle, 2002 ) and integrate disparate concepts (Cowan etal., 2005 ; Oberauer etal., 2007 ). These attention and memory processes, in turn, facilitate reasoning processes. More simply, people with high working memory capacity are at a particular advantage when working through novel problems, since they can consider more information at any given moment.
Unfortunately, if one desires to truly understand the nature of human intelligence, the working- memory- as- a- cause explanation is rather limited. All it really states is that memory is a prerequisite to reasoning and that tests of working memory capacity are particularly good measures of this type of memory.
Nonetheless, we do believe that working memory is critical to human reasoning. However, our work has moved in a direction where we now assume that classic tests of working memory capacity provide only a limited view of the working memory system as a whole. In particular, we believe that individual differences in fluid intelligence can provide unique insights into the qualitative interactions between attention and memory.
These insights advance both our understanding of the working memory system and its contributions to complex thought.
Working memory capacity tasks are, by their nature, focused on the aspects of working memory that allow a person to maintain access to relevant information (either through active maintenance or through retrieval; Unsworth & Engle, 2007 ). Yet, in a true “working” memory system, the relevancy of information will be subject to constant change. This system will therefore also require mechanisms of disengagement (e.g., inhibition, episodic tagging, unbinding of temporary associations) that allow for the removal of no- longer- relevant information (e.g., updating; Ecker et al., 2010 ; Miyake et al.,2000 ). These disengagement mechanisms (Figure 18.1), however, are not likely to be critical to the performance of most standard measures of working memory, since these tasks require people to retain access to information. Forgetting of information will lead to lower test scores, and thus be deemed an indicator of low working memory capacity. In all likelihood,
Our contention is that disengagement is the factor that accounts for the less than perfect correlation between working memory capacity and fluid intelligence. While forgetting is detrimental to performing a working memory task, it is advantageous to tests of fluid intelligence. When reasoning through a problem, people will have mistaken assumptions that can lead to dead- end thinking. Disengagement allows for a fresh perspective and for blocking outdated assumptions.
From this perspective, we do not see the factors underlying working memory capacity and fluid intelligence as representing different cognitive systems (Engle, 2002 ). Instead, we view both as arising from the working memory system (see bottom of Figure 18.1 ).
Working memory capacity and fluid intelligence task performance correlate due to the common need to engage executive attention, in order to define and carry out goals. Both maintenance and disengagement generally require executive attention. The less than perfect correlation between working memory capacity and fluid intelligence task performance is attributable to the tests being differentially sensitive to critical working memory functions. Th at is, executive attention engages different cognitive mechanisms when these tasks are completed.
Our perspective makes a simple, straightforward prediction. If tests of fluid intelligence indeed tap into the disengagement- related aspects of working memory, then individual differences in fluid intelligence should reflect the ability to forget outdated information in simple memory tasks (the type that require very little reasoning). Moreover, this aspect of reorganizing the contents of working memory will not be attributable to individual differences in working memory capacity (at least not as they are traditionally defined). To be clear, our argument is not (working memory capacity = maintenance) and (fl uid intelligence = disengagement). It is that tests of working memory capacity place primary demand on the maintenance- related aspects of the working memory system and secondary demand on the disengagement aspects. Fluid intelligence tests place demand in the reverse order. Th us, the observed constructs of working memory capacity and fluid intelligence do not represent separate cognitive abilities, so much as they represent different and sometimes contradictory functions that are necessary for information processing.
Th ese different degrees of reliance lead the constructs to be somewhat unrelated (and even contradictory). But working memory tasks do reflect between- trials disengagement (e.g., Kane & Engle, 2000 ) and fluid intelligence tasks do require mental representations to be maintained. This leads to a degree of relation. More importantly, maintenance and disengagement processes are organized around goals, and subject to attentional resources, which further strengthens the relation between these varieties of task. From this perspective, traditional measurement of a person’s working memory capacity (using complex span or visual arrays tasks) has provided a limited view of the functions carried out by working memory. This is because the primary goal in these tasks is to remember as much information as possible. Thus, traditional measurement of working memory