Taking visual word recognition as a paradigm case for investigating mental function, cognitive neuroscience has made progress in understanding this ability as coordinating three mental codes: visual, phonological, and semantic (Posner 1978). Progress has been made in associating these codes with activity in particular brain regions. (Posner and McCandliss 1999). We focus here on two areas that have proven importance for understanding individual differences in reading ability.

A phonological region of interest (Phonological -ROI). Several converging forms of evidence suggest a link between phonological processing ability and a left posterior dorsal brain region (left parieto-temporal-see figure 1) which includes the posterior extent of the superior temporal gyrus, angular gyrus, and supramarginal gyrus. Supporting evidence includes lesion studies that link specific patterns of physical damage to specific losses of function (Warrington and Shallice 1980), as well as by neuroimaging studies that manipulate phonological task demands and show a systematic effect on hemodynamic activity (Carr and Posner 1995),(Fiez and Petersen 1998)). The relation between phonological demands and region activity also helps account for individual differences; adults who are poorly skilled or impaired on phonological awareness fail to show the normal pattern of increased activity in this region under phonologically demanding tasks such as word and non-word rhyming, under visual or auditory presentation (Paulesu, Frith et al. 1996; Salmelin, Service et al. 1996; Rumsey, Horwitz et al. 1997; Shaywitz, Shaywitz et al. 1998). Such findings have recently been extended to children, demonstrating these same structure- function relationships and individual differences during the early years of reading acquisition (McCandliss 2001; Temple, Poldrack et al. 2001). Finally, several recent studies have demonstrated that for children who have difficulties in reading and phonological awareness, brain imaging measures applied before and after intensive phonological interventions lead to increased recruitment of this region during phonologically demanding tasks. (McCandliss 2001; Simos, Fletcher et al. 2002; Simos, Fletcher et al. 2002; Temple, Deutsch et al. 2003). These studies converge on the finding that instructional interventions that lead to increased phonological awareness lead to changes in brain activity -- the phonological region of interest is recruited to a greater degree during phonologically demanding tasks. It should be noted, however, each of these intervention studies are limited by the fact that the learning was not contrasted to another form of instruction, but rather to a no-intervention control.

An alphabetic symbol region of interest (Alphabetic ROI). A similarly compelling case can be made for link between visual codes for the letters within words, and a left ventral occipito-temporal visual system, near the inferior temporal/fusiform gyrus (see figure 1). (Fiez and Petersen 1998; McCandliss in press). Physical damage to this region damages the ability to read words, but not to hear them. (Binder and Mohr 1992; Cohen, Martinaud et al. in press). Neuroimaging studies show this region to have a greater responses to visual words than to auditory words, pseudo-letters, or checkerboard patterns. (Cohen, Dehaene et al. 2000). Subtle individual differences in letter recognition are correlated with activation levels in adults (Garrett, Flowers et al. 2000), and poorly-skilled readers in many languages demonstrate decreased activity in this region compared to skilled readers (Paulesu, Demonet et al. 2001). This structure function association has recently been extended to explain individual differences in children’s reading abilities, demonstrating a positive correlation between ability and activity in this region when processing written word stimuli (Shaywitz, Shaywitz et al. 2002) . Note that correlations between individual differences in alphabetic ability are not driven only by reading impairments, but are significant within the range of poor readers and good readers alike. Finally, there is some preliminary evidence that activity in this region can be influenced by instructional intervention. Intervention studies that happen to combine phonological instruction with letter-sound mapping activities demonstrated some intervention-related changes in this ventral region of interest. The proper controls to isolate this possibility, however, have not been run to date.

The ‘core code’ instruction and brain activity. Several prominent theories hold that difficulty in phonological awareness is the fundamental difficulty in underachieving readers, and this ‘core code’ might be targeted best through verbal, acoustic, phonological awareness activities ((Fletcher 1994) (Merzenich 1996)) (Lovett 1989). This approach would hold that focused instruction of this type will strengthen the tendency to activate the phonological ROI, thereby removing the core ‘deficit’ that was the source of difficulty.

The ‘integrated codes’ instruction and brain activity. A complementary approach holds that the typical course of learning to read involves integrating visual letter symbol and phonological codes (Perfetti 1985) (Share and Stanovich 1995) (Cunningham, Perry et al. 2002), and this is achieved through focusing attention on letter-sound relationships as words are being decoded (McCandliss, Sandak et al. 2003). Although low phonological awareness might play a role in preventing this process from occurring, interventions should focus directly on strengthening the association between letters (groups of letters) and their sounds (phonological codes). Such activities may increase co-activation of the Alphabetic ROI and the Phonological ROI and during decoding activities. In addition, it has often been proposed that instructional focus on letter-sound relationships within words may help advance phonological awareness skills, by drawing attention to subtle phonological contrasts (Morais 1979) (Perfetti 1985). Perhaps, brining two integrated systems (alphabetic and phonological) to bear on a problem improves processing ability (Ehri 1999).

The last decade of cognitive neuroscience research has also provided evidence from multiple convergent sources that numbers are represented via the coordination of three distinct mental codes: visual symbols, verbal labels, and analog-magnitude representations (Dehaene, 1992), each of which can be linked to dissociable brain regions (Dehaene, Piazza et al. 2003).

The analog-magnitude code is hypothesized to function as an “internal number line” (Dehaene, 1992), which allows people to make decisions about relative sizes of numbers, estimate the results of calculations, and understand the meaning of arithmetic operations. In other words, it provides “number sense” by relating each number symbol to its significance. Although linking digit symbols with verbal labels rarely poses a persistent challenge to learning, the ability to link number symbol and magnitude has been argued to be a gateway skill that poorly performing children are lacking (Griffin, Case et al. 1994) (Griffin, Case et al. 1995)

An analog-magnitude brain region of interest (Magnitude ROI). A recent meta-analysis of multiple sources of evidence, creates a compelling case that analog-magnitude codes critically involve an area of the right parietal cortex called the horizontal intra-parietal sulcus (HIPS), (see figure 2). (Dehaene, Piazza et al. 2003). Damage to this region can impair magnitude judgments while sparing number symbol and word abilities (e.g. Dehaene & Cohen, 1997). This region has been directly implicated in neuroimaging contrasts that manipulate challenging versus easy magnitude comparison tasks while controlling for symbolic codes. (e.g. Dehaene, 1999; Stanescu- Cosson et al., 2000)—a finding that has recently been replicated in children(Temple and Posner 1998) . Activation in this region has recently been linked, in adults, to individual differences in numerical processing (Dehaene, Cohen et al. 1998), and to number processing impairments in children (Isaacs, Edmonds et al. 2001). Unlike the Phonological ROI, the responsiveness of this region to targeted instruction is not yet fully established, but it remains a fruitful possibility, as research on number skill intervention has tended to lag behind reading by several years (Ansari & Karmiloff-Smith, 2002), despite the many parallels.

Number-symbol brain region of interest. Dehaene and colleagues have hypothesized that Arabic digits are represented separately from other quantity codes, along left and right ventral occipito-temporal pathways (Chochon 1999) (see figure 2.) Some patients with brain damage to these regions are unable to recognize digits, yet retain mental calculation and some other visual abilities (Cohen and Dehaene 1995). Although such regions are similar in several ways to the Alphabetic ROI described above, regions for number symbols are distinct from alphabetic symbols, as revealed by distinct patterns of activation in fMRI studies with adults (Polk, Stallcup et al. 2002). (Pinel, Le Clec et al. 1999; Pinel, Dehaene et al. 2001). Such findings in adults have recently been extended to children via an electrophysiological study demonstrating distinct responses between quantities presented as collections of dots versus Arabic symbols, which in turn were demonstrated to be separate from magnitude effects (Temple and Posner 1998). These findings demonstrate that different brain systems can be identified that distinguish between processing of alphabetic and numerical symbols, and that such visual symbol processing can be dissociated from more conceptual processes such as number magnitude values.

Instructional Approaches. Like word reading, early instruction in number exhibits a tension between core and integrated code approaches. It is a common practice to have children work with perceptual magnitudes. For example, children may physically manipulate quantities to make judgments of larger and smaller, and they may map the physical quantities and their own actions to symbolic notations and procedures While it seems self-evident that children need to ground mathematical notations in perceived magnitudes, there is little agreement on how and when the manipulation of perceptual quantities helps (Chao et al., 2000, Uttal et al., 1997). For example, some assume that the quantitative meanings of perceptual magnitudes are self-evident (Cary & Carlson, 1999), and therefore, training designed to help children notice perceptual differences in quantity would be unnecessary. Alternatively, a core code approach would suggest it is necessary for children to learn to perceive the relevant perceptual quantities and their relations, for example, by helping children make perceptual comparisons of near magnitudes (Bransford et al., 2000). The integrated code alternative would ask children link symbolic notations and perceptual magnitudes so that the symbolic notations help children learn to organize perceptual quantities into useful structures while also motivating the meaning of the symbolic notations (Moreno & Mayer, 1999; Schwartz & Moore, 1998). Case & Okamoto (1996), for example, propose that the mental number line develops through the integration of perceived quantities, counting words, and pointing. Case and others (Griffin, Case & Siegler, 1994; Griffin, Case, & Capodilupo, 1995) designed an intervention to target the development of the internal number line, known as “RightStart.” Kindergarten children at risk for mathematical difficulties participated in group games which encouraged their understanding of the basic principles of the number line. Children who completed RightStart showed large gains in number sense compared to control children, and these gains helped them succeed in arithmetic classes the following year. In Case’s study, however, it is not clear whether the cause of the gains was primarily through remediating deficits in the visuo-spatial quantity system or whether it was also through the strengthening of connections between the digit and quantity systems. To tease apart these alternatives, it is necessary to compare core code instruction that targets quantity differentiation and visuo-spatial development with integrated code instruction that also emphasizes linking the visuo-spatial and symbol systems.

The proposed experiments will compare effects of reading and number training using the contrastive instructional treatments of perceptual training to no-training (Exp. 1) and perceptual training to perceptual- symbol linking (Exp. 2). Instructional software will implement the differences in treatments by varying theoretically critical elements. By using software as the main instructional vehicle, we do not mean to imply that the ultimate goal for instructional software is to be a stand-alone application, especially for the young children in these studies. Instead, for the current purposes, it provides a way to implement critical experimental contrasts while minimizing the potential confounds of strong situational learning factors. The software will leverage current NSF funding and prior work by the PIs that has successfully used contrasting cases and Teachable Agents as pedagogical means.

Table 1. Illustration of instructional contrasts.

	SKILL DOMAIN
	Literacy	Numeracy
Strengthen Core Code	Verbal Phonological Comparisons (Phonological Awareness)	Quantity & Magnitude Comparisons (Analog Estimation Skills)
Integrate Codes	Letter-sound Associations (Decoding Skill)	Symbol-Magnitude Association (Number Sense)