The neural substrate for concrete, abstract, and emotional word lexica: a positron emission tomography study.

INTRODUCTION

Viewing written words activates a set of abstract codes in the brain-visual codes for perceptual analysis, an orthographic code that represents the visual form of known words, codes for the word's meaning (stored in semantic memory), and a code for its sound-based (phonological) output form (Posner & Carr, 1992). The exact nature of, and relationship between, these codes remains a matter of considerable controversy (Rumelhart et al., 1986; Shallice, 1988). Detailed case reports of brain-damaged patients have demonstrated selective impairments to specific components of the reading system, providing striking evidence for a basic modular organization of reading processes (Shallice, 1988). At the same time, imaging studies, generally with positron emission tomography (PET), have provided parallel evidence that many "elementary operations forming the basis of cognitive analysis of human tasks are strictly localized" (Posner, Petersen, Fox, & Raichle, 1988, p. 1627). This paper reports evidence linking activation of one of the cognitive components of reading - the orthographic lexicon - to regions of the left temporal lobe when different types of written words are presented for passive viewing. We will also suggest that further limbic areas of the brain are engaged when the material is emotionally salient.

Evidence from cognitive studies and neuroimaging have tended strongly to support the notion of strict localization of the visual word form, henceforth the orthographic lexicon, in the human brain. Considerable controversy continues to exist as to whether it is localized in the left occipital lobe or, alternately, in the left temporal lobe.

Cognitive and neuropsychological evidence has pointed in both directions. Damage to the left occipital lobe and posterior corpus callosum produces a pure reading disorder (alexia without agraphia, sometimes termed pure alexia), in which subjects are only able to read words laboriously in a letter-by-letter fashion if at all. This alexia is dissociated from other language deficits, and this has suggested to other investigators that the left occipital region translates letter elements into familiar orthographic patterns that represent word forms (Damasio & Damasio, 1983; Geschwind, 1965). Indeed, the view that the stored form of words might be localized in the left occipital lobe remains the majority position among most clinicians in neuropsychology.

Recent evidence from detailed cognitive studies of pure alexia following damage to extrastriate cortex, however, disputes this interpretation of the syndrome of alexia without agraphia. Patients with left extrastriate lesions and pure alexia still identify individual letters faster and more accurately in a word than in a random-letter array (Bowers, Arguin, & Bub, 1996; Bowers, Bub, & Arguin, 1996; Bub, Black, & Howell, 1989; Reuter-Lorenz & Brunn, 1990), even when guessing is prevented by stringent forced-choice procedures. This indicates strongly that there is a continuing effect of orthographic structure on word perception in such patients (Johnston & McClelland, 1980). Furthermore, some pure alexics exhibit covert reading ability at exposures brief enough to preclude explicit identification of the target items (Bowers, Arguin, & Bub, 1996; Bowers, Bub, & Arguin, 1996; Coslett & Saffran, 1989; Shallice & Saffran, 1986). These data suggest that patients with left occipital lesions and alexia continue to have a functional orthographic word store.

If the orthographic lexicon is not localized in the left occipital area, where might it reside? An alternative proposed localization for the orthographic lexicon is in the left posterior temporal lobe region. The argument in favor of this region runs as follows: pure alexia may represent the wrong target population to study in order to delineate the neural substrate of the orthographic lexicon. The two-route model for reading words in English (Patterson, Coltheart, & Marshall, 1985) implies that even with impaired access to the lexicon, regular words should be read well using spelling-to-sound correspondences. Patients with damage to the orthographic lexicon should preferentially be affected in their reading of irregular words - words like pint (c.f. hint, mint, stint, etc.) that violate the regular correspondence between spelling pattern and sound. While pure alexia patients fail to display such a pattern of deficits, other brain-damaged patients do exist who show this pattern, and they are termed surface dyslexics. These patients may read at normal speeds but treat many common words as novel spelling patterns having no stored representation (Bub, Cancelliere, & Kertesz, 1985; Patterson et al., 1985). A review of neuroanatomical correlates of surface dyslexia concluded that the critical lesion sites include temporal lobe structures, and in a number of cases the damage was largely restricted to the left temporal lobe (Vanier & Caplan, 1985). The case may thus be made for the left temporal lobe as the neural substrate of the orthographic lexicon.

Studies of functional brain imaging using PET provide another approach to localizing the orthographic lexicon in normal individuals (Chertkow & Bub, 1994; Demonet, Wise, & Frackowiak, 1993; Posner & Carr, 1992). PET studies of single-word processing in English, however, have similarly produced a conflicting set of results taken as supporting both the occipital and the left temporal lobe locus for the orthographic lexicon.

In seminal PET studies, Petersen, Posner, and colleagues proposed that the orthographic lexicon is indeed localized in the left extrastriate region, since silent viewing of words compared to a "flashing plus sign" baseline condition produced activation involving the occipital lobe but not temporal lobe cortex (Petersen, Fox, Posner, Mintun, & Raichle, 1988; Petersen, Fox, Snyder, & Raichle, 1990; Posner, Petersen, Fox, & Raichle, 1988; Posner & Petersen, 1990). In their first study (Petersen et al., 1988), presentation of concrete words (animal names) for a duration of 150 msec, with one word presented every second, produced activity in both left and right striate cortex. In addition, activation was produced in left and right "extrastriate cortex" at 2.4 cm from the midline, just above the level of the intercommissural line.

In the second study (Petersen et al., 1990), a similar presentation of mixed animals and household object words was carried out, along with a condition of false fonts (a series of abstract designs with structural similarity to letters), letter strings (all consonants), and pronounceable nonwords. The goal of using these other stimuli was to allow better separation of activations related to visual as opposed to phonological or semantic processing. All of the stimuli produced activity in "lateral extrastriate regions" when compared with a plus sign baseline condition. Real words and nonwords produced activity in the same left medial extrastriate region that was activated in the first study. This was accompanied, for real words only, by activation of the left inferior frontal region. These two studies were taken as evidence that orthographic encoding was largely localized to the left medial extrastriate region, inasmuch as viewing of consonant strings and false fonts (which were unlikely to activate either orthography or phonology) failed to activate this area. Furthermore, the fact that meaningful words, but not other "word-like" stimuli, produced activation visible in the left inferior frontal region suggested a semantic encoding role for this frontal area. Indeed, these studies had identified activation in this same region only during semantic processing tasks. The left temporal lobe was notable for its absence in these experiments.

In contrast, the study of Howard, Patterson, Wise, Brown, Friston, Weiller, and Frackowiak (1992) utilized reading words aloud (exposure duration 1000 msec, with presentation of one word per 1.5 sec) and oral repetition, compared with visual and auditory repetition baselines. These two conditions produced activation of the left posterior superior temporal gyrus. Their study utilized words that were a mixture of abstract and concrete items. This activation for visual word reading was 1.6 cm away from the peak of activation produced by auditory presentation of the same words. There was no activation in the medial left extrastriate cortex. A follow-up study by Price, Wise, Watson, Patterson, Howard, and Frackowiak (1994) examined both reading aloud and silent viewing of words, the latter carried out both at a short-exposure duration (150 msec), akin to Petersen et al. (1990), and a long exposure duration akin to Howard et al. (1992). The baseline condition consisted of viewing false fonts only. In silent viewing of words compared to false fonts, they reported activation of the "bilateral inferior occipital cortex," particularly significant at the longer exposure duration. This Price and colleagues ascribed to "greater visual stimulation" (1994, p. 1265). They noted that "activity in the medial extrastriate cortex did not reach significance at either exposure duration" (p. 1265). Most strikingly, their study produced robust activation of the left posterior middle temporal gyrus, both at short- and long-exposure durations for words. This left temporal area was about 8 mm posterior to the area activated by visual words in Howard et al. (1992). The divergence between these two sets of results is striking.

A further study by Bookheimer, Zeffiro, Blaxton, Gaillard, and Theodore (1995) examined reading written words silently (concrete nouns) as well as reading them aloud and compared these conditions to naming of objects. The comparisons were with a baseline consisting of silently viewing complex, meaningless figures made up of line drawings. In their silent viewing of written words condition, they failed to find activation in the posterior middle temporal gyrus. Instead, they demonstrated activation in the left posterior inferior temporal region and the nearby fusiform gyrus, adjacent to the "basal temporal language area" (Burnstine et al., 1990; Luders, 1991). Additionally, the …