Processing efficiency in preschoolers’ memory span: Individual differences related to age and anxiety

Abstract

In self-paced auditory memory span tasks, the microanalysis of response timing measures represents a developmentally sensitive measure, providing insights into the development of distinct processing rates during recall performance. The current study first examined the effects of age and trait anxiety on span accuracy (effectiveness) and response timing (efficiency) measures from word and digit span performance in a preschool sample (N = 76, mean age = 57 months, SD = 11). Children were reassessed 8 months later using the same two tasks plus a test of nonword memory span and a measure of articulation rate. The results at the second time point (T2) confirmed the effects of age on both processing effectiveness and efficiency. Trait anxiety was an additional negative predictor of span effectiveness (especially for digit span) and efficiency (in the case of word and nonword span). The findings are discussed in the context of factors contributing to early short-term memory development and attentional control theory.

Introduction

Short-term memory (STM) represents the ability to generate and store mental tokens that code for entities in the environment (Feigenson, 2007). Its development has been studied extensively with a recent focus on its emergence during infancy and early childhood (Oakes & Bauer, 2007). The memory span procedure measures how many items a person can repeat back in sequence and has been widely used in adult and developmental studies as an important tool for investigating STM development. This procedure has a number of important advantages. First, it is simple enough to be understood by children as young as 2 years 10 months (Gathercole & Adams, 1993). Second, it is a developmentally sensitive index considering that it increases steadily between 3 and 10 years of age (Alloway et al., 2006, Dempster, 1985, Gathercole and Adams, 1993). Finally, research into the mechanisms involved in performing this task is ecologically relevant and has been related to comprehension and problem-solving abilities (Dempster, 1985), academic achievement (Alloway et al., 2005, Swanson, 1994), and general intelligence (Hutton & Towse, 2001).

A promising line of research investigating the development of auditory memory span was proposed by Cowan and colleagues (Cowan et al., 1992, Cowan et al., 1994, Cowan et al., 1998, Cowan et al., 2003, Cowan et al., 2006, Cowan et al., 1999, Hulme et al., 1999). By focusing on the self-paced nature of the verbal output, this research revealed the distinct informative value of different response timing measures, such as preparatory intervals, interword pauses, and word durations, in explaining the development of memory span performance. This provided an alternative account to the hypothesis that articulation rate, as an index of subvocal rehearsal, offers the best estimate of memory span. Developmental and adult evidence for a linear relation between articulation rate and verbal span has supported the latter hypothesis (e.g., Baddeley et al., 1975, Cowan et al., 1992, Gathercole and Adams, 1993, Henry, 1994, Hitch and Halliday, 1983), although the relation was not always straightforward (Hitch et al., 1993, Hulme and Muir, 1985). In addition, in the developmental context, measures of articulation rate have been shown to be themselves “contaminated” by memory effects and largely redundant with age in explaining variance in memory span (Ferguson, Bowey, & Tilley, 2002).

The microanalysis of response timing patterns also refined the more general account based on a global development in processing speed (Cerella and Hale, 1994, Kail, 1992, Kail and Salthouse, 1994, Salthouse, 1996). The global processing speed account, revised by Kail (1997), was recently challenged by the results of Ferguson and Bowey (2005), whose findings suggested that instead of being directly connected to memory span, “processing speed largely mediates developmental increases in some underlying construct that, in turn, influences memory span” (p. 108). However, the underlying mechanisms behind this mediation remain unclear.

Cowan and colleagues made several inferences regarding the underlying mechanisms of STM functioning from the microanalysis of response timing measures during overt recall. These indexes were analyzed concurrently with an articulation rate measure in both STM tasks (e.g., Cowan et al., 1994, Cowan et al., 1998, Hulme et al., 1999) and working memory (WM) tasks (Cowan et al., 2003). Several findings emerged as a result of these studies, which examined the interrelationships among age, response timing, and memory span. First, the speed of distinct output elements (especially interword pauses and preparatory intervals) represented reliable independent predictors of memory span performance. Second, interword pauses were affected by list length but unaffected by word length (Cowan et al., 1994). Third, in adults, interword pauses were much longer for nonwords than for words (Hulme et al., 1999). Fourth, interword pauses, preparatory intervals, and word durations were sensitive to individual differences in age and memory performance, so that (a) with age, and for a particular list length, all of these indexes decreased, suggesting more efficient processing; (b) speeded-speaking measures predicted span in first graders but not in older children, whereas interword pause durations predicted span in older children but not in younger children; and (c) word durations and interword pauses were shorter within children of a particular age who had higher memory spans (Cowan, 1992, Cowan et al., 1998). Therefore, response timing measures (beyond pure speeded-speaking estimates such as word durations) were shown to be sensitive to individual differences in age and memory performance. Because they were also influenced by the memory demands of the task, they represent an ideal candidate for the investigation of memory span development.

The current study investigated the mechanisms involved in the development of auditory memory span in preschoolers advocating the relevance of response timing measures in the analysis of STM processes. This age group is underresearched given that it might yield important insights into the emergence and early functioning of serial recall processes in children. There is evidence to show that the preschool years represent a time of intensive development in memory strategies, characterized by large individual variations in memory strategy (Schneider, Kron, Hűnnerkopf, & Krajewski, 2004; Sodian & Schneider, 1999). However, to our knowledge, no study so far has investigated the development of response timing patterns and their relation to memory span in preschoolers within a longitudinal framework.

Individual variations in distinct response timing segments in this age group could be affected by trait anxiety. The most influential explanatory model regarding the anxiety–WM relationship is the processing efficiency theory (PET) (Eysenck & Calvo, 1992) or, in its revised form, the attentional control theory (ACT) (Derakshan and Eysenck, in press, Eysenck et al., 2007). At the core of the PET was the distinction between processing effectiveness (quality of task performance indexed by standard behavioral measures, i.e., accuracy) and processing efficiency (time taken to complete the task, subjective mental effort). The theory predicted that in high-anxious individuals, efficiency decreased as more effort and resources were invested to complete a task or to attain a given performance level because of concurrent resources being consumed by anxious thoughts (worries). Within the classical WM model and its subsequent revisions (Baddeley, 1986, Baddeley, 2000, Baddeley, 2007), the detrimental effects of anxiety were predicted to appear more at the level of the central executive, impairing attentional control, and of the phonological loop (Eysenck et al., 2007). The adverse effects of anxiety on performance efficiency become greater as overall task demands on the central executive increase (Eysenck et al., 2007). Research with adults has shown that high- and low-anxious individuals have comparable performance effectiveness but differ in terms of performance efficiency (Ansari et al., 2008, Derakshan et al., in press, Derakshan and Eysenck, 1998; Ikeda, Iwanaga, & Seiwa, 1996; Markham & Darke, 1991).

Few studies have investigated the anxiety–memory relationship during early development, and there is no conclusive evidence concerning trait anxiety and its impact on processing efficiency on verbal STM span tasks in children (see Visu-Petra, Ciairano, & Miclea, 2006, for a review). With regard to this developmental context, Daleiden (1998) advocated that global approaches, emphasizing only anxiety-related attentional biases during the encoding stage, have been inconclusive in explaining anxiety effects on general memory performance. Such global approaches imply the presence of deficits in the initial attentional processes that interfere with the representation of the information in memory (Derryberry & Reed, 1996). Indeed, while solving a cognitive task, children with high anxiety were shown to produce significantly more task-inhibiting thoughts and ongoing negative evaluations than were those with low anxiety (Francis, 1988). Other studies describe difficulties in shifting of attention from internal to external stimuli (Kendall & Chansky, 1991).

Taking into account both effectiveness and efficiency indexes of memory performance, one study by Hadwin, Brogan, and Stevenson (2005) attempted to directly investigate the relationship between state anxiety and STM and WM in older children (9–10 years) within the explanatory framework of the PET. State anxiety (measured with the State–Trait Anxiety Inventory for Children [Spielberger, 1973], a self-report measure for children between 9 and 12 years of age) did not influence performance accuracy on the verbal memory task. However, it had a significant impact on the total time taken to complete the verbal memory tests and on self-reports of mental effort, both measures of distinct facets of processing efficiency.

The general aim of the current investigation was to replicate and extend the previously mentioned findings in several directions to improve our understanding of STM development during the preschool years. The main extensions consisted in the longitudinal nature of the study, the range of measures involved, and the relation to individual differences in trait anxiety. First, considering the developmental trends issue, a previous detailed analysis of response timing patterns in very young children has been performed for only a relatively small (n = 16) group of children, with a mean age of 4 years 5 months (Cowan et al., 1994), assessed at a single time point. Other previous studies with children have either compared preschoolers with much older children (4 years 5 months vs. 8 years 8 months [Cowan et al., 1994]) or compared school-aged children from different grades (first, third and fifth grades [Cowan et al., 1998, Experiment 1]). In the current longitudinal study, at the first time point (T1), the younger participant group (n = 39) represented a broader age range from 37 to 59 months with a mean age of 4 years (SD = 8 months); they were cross-sectionally contrasted with an older age group (n = 37, age range from 60 to 74 months, mean age 5 years 6 months [SD = 4 months]). Both age groups were reassessed 8 months later. This approach could reveal smooth developmental transitions that would gradually approximate the pattern seen in older children. However, it is equally possible that a different pattern of response timing might appear in preschoolers’ verbal output measures, revealing that particular memory spans are accompanied by different profiles of processing rates in children of different ages (Cowan, 1999).

A second extension of previous research was related to the measures included in this study. At the first time point, to investigate whether the response timing patterns were similar irrespective of stimulus type (Gathercole & Adams, 1993), we used both digit and word span measures. At this initial stage, parental reports of the children’s anxiety levels were also taken using the Spence Preschool Anxiety Scale (Spence, Rapee, McDonald, & Ingram, 2001). At the second time point (T2), additional measures of nonword recall and articulation rate were introduced. Although the articulation rate measure has been widely used in relation to the response timing measures, there were three main reasons for including the nonword memory span task. First, Gathercole and Adams (1993) showed that it could be used with children as young as 2 years 10 months. Moreover, in the same study, nonword memory span was significantly associated with articulation rate, unlike digit span, probably because the former was more constrained by children’s speech production skills than was the latter. Second, nonword memory span is considered as a particularly useful measure of children’s STM abilities (Gathercole & Baddeley, 1989) because it provides a “purer” measure of phonological capacity without the confounds from long-term memory generated by the stimuli used in word or digit span tasks. Third, adult findings of within-participant relationships between measures of memory span and response timing vary according to the lexical status of items to be remembered, with shorter interword pauses for words than for nonwords (Hulme et al. 1999).

Extending the predictions of the ACT (Eysenck et al., 2007) to this age range, we decided to use the microanalysis of response timing measures to compare performance effectiveness and efficiency at different levels of trait anxiety. State anxiety could not be reliably assessed via self-reports at this young age, so parental ratings of child anxiety were used as a measure of trait anxiety (Spence et al., 2001). PET advocates that even when low- and high-anxious individuals have comparable performance effectiveness, group differences in efficiency can be inferred from differences in response time (Eysenck, Derakshan, Santos, and Calvo, 2007). We considered that a global measure of total response time (not differentiated across verbal output elements or across list lengths) used by Hadwin and colleagues (2005) might obscure the impact of anxiety on the specific processing that occurs during distinct response segments. Therefore, a final main aim of this study was to identify the concurrent and predictive effects of preschool anxiety on both performance effectiveness and efficiency indexes measured from the three types of span measures (digit, word, and nonword span).

Considering these objectives, we formulated the following predictions regarding, first, the effects of age and task demands and, second, the underresearched effects of trait anxiety on memory performance. First, we predicted that (a) the duration of all response timing indexes would decrease with age (from both cross-sectional and longitudinal perspectives) and increase with list length; (b) the durations of preparatory intervals and of interword pauses (but not of the words themselves) would represent independent predictors of span performance on the word, digit, and nonword span measures; and (c) conversely, span performance itself would represent a determinant of response timing measures. Second, our predictions regarding trait anxiety were that (a) trait anxiety level will not represent an independent predictor of span accuracy, but (b) the predictive negative effect of anxiety levels will be visible when the “strategic” processing efficiency elements, such as interword pauses and preparatory intervals, are considered, as compared with word durations.