Neural responses to morphological, syntactic, and semantic properties of single words: An fMRI study

Abstract

Dissociations in the recognition of specific classes of words have been documented in brain-injured populations. These include deficits in the recognition and production of morphologically complex words as well as impairments specific to particular syntactic classes such as verbs. However, functional imaging evidence for distinctions among the neural systems underlying these dissociations has been inconclusive. We explored the neural systems involved in processing different word classes in a functional Magnetic Resonance Imaging study, contrasting four groups of words co-varying morphological complexity (simple, monomorphemic words vs complex derived or inflected words) and syntactic class (verbs vs nouns/adjectives). Subtraction of word from letter string processing showed activation in left frontal and temporal lobe regions consistent with prior studies of visual word processing. No differences were observed for morphologically complex and simple words, despite adequate power to detect stimulus specific effects. A region of posterior left middle temporal gyrus showed significantly increased activation for verbs. Post hoc analyses showed that this elevated activation could also be related to semantic properties of the stimulus items (verbs have stronger action associations than nouns, and action association is correlated with activation). Results suggest that semantic as well as syntactic factors should be considered when assessing the neural systems involved in single word comprehension.

Introduction

An important question in understanding the cognitive and neural basis of language processing concerns the representation of different word types in the brain. Recent years have seen evidence for a number of neuropsychological dissociations that suggest neural separation of the systems involved in processing different classes of words. For instance, considerable evidence now suggests that nouns and verbs can be differentially impaired follow brain injury (see for instance Caramazza & Hillis, 1991; Damasio & Tranel, 1993). Other dissociations have been reported between tasks involving morphological simple words (like jump or dark) and words composed of multiple morphemes (such as jumped, darkness Marslen-Wilson & Tyler, 1997; Marangolo et al., 2003).

One interpretation of these dissociations is that this neural separation reflects core linguistic distinctions between mechanisms of word storage and grammatical processing (Pinker, 1999; Ullman, 2001). For instance, dissociations between verbs and nouns may arise as a consequence of the greater complexity of the grammatical information that is required for verbs to be used (Shapiro & Caramazza, 2003). The comprehension of verbs may therefore place a greater demand on neural systems involved in grammatical operations than do nouns, hence the vulnerability of verbs following brain injury reflects specialisation for grammatical processing operations. Similarly, since polymorphemic words may be decomposed during identification (Caramazza, Laudanna, & Romani, 1988; Marslen-Wilson, Tyler, Waksler, & Older, 1994) additional processing operations for decomposition may occupy a distinct neural locus, damage to which may impact on the recognition of polymorphemic words (Tyler et al., 2002; Ullman et al., 1997).

An alternative account of these dissociations suggests that these deficits are a consequence of the different phonological and semantic properties of these classes of words. For instance, nouns tend to have more concrete meanings than verbs. An apparent verb impairment may arise from a greater difficulty in processing abstract words following brain injury (Bird, Howard, & Franklin, 2000), though this remains controversial (Shapiro & Caramazza, 2003). In accounting for dissociations involving morphologically complex words it has been proposed that regularly inflected words might depend more strongly on systems involved in phonological processing—for instance, due to the greater phonetic complexity of inflected forms (Bird, Lambon Ralph, Seidenberg, McClelland, & Patterson, 2003). These proposals therefore suggest an alternative, non-grammatical account of impaired performance on verbs and morphologically complex words in brain-injured patients.

Functional neuroimaging investigations have an important role to play in assessing these contrasting explanations of the representation and processing of different word types. However, a flurry of work in recent years has not led to any consistent conclusions concerning the neural systems that are differentially activated by nouns and verbs. For instance, while some studies have reported regions of distinct activation in the posterior middle temporal lobe for verbs compared to nouns (Kable, Lease-Spellmeyer, & Chatterjee, 2002; Perani et al., 1999) other studies have failed to replicate these results (Tyler, Russell, Fadili, & Moss, 2001). One possible factor in explaining these discrepancies may be a linguistic difference between stimuli in languages such as Italian (as in the Perani study) that are marked with inflectional endings specific to verbs and stimuli in English which are typically uninflected and are therefore ambiguous between nouns and verbs (for instance, the unmarked stem walk may be either a noun or a verb).

Prior neuroimaging work that has investigated the processing of morphologically complex words (Beretta et al., 2003; Jaeger et al., 1996; Marangolo, Piras, Galati, & Burani, 2003) have focussed on assessing the neural correlates of the overt morphological operations involved in producing complex forms from bare stems (though see Laine, Rinne, Krause, Teras, & Sipila, 1999). These studies are of interest in demonstrating neural systems that may be involved in performing these morphological operations. However, these studies are not sufficient for us to conclude that these same mechanisms of overt morphological manipulation are damaged in patients that show a deficit in processing morphologically complex words. Evidence for neuropsychological deficits on complex words can be observed in tasks such as reading aloud (Rastle, Tyler, & Marslen-Wilson, 1999), semantic (Longworth, 2002) or repetition priming (Marslen-Wilson & Tyler, 1997) that do not necessarily involve overt morphological manipulation. It is therefore unclear whether the neural systems activated in these neuroimaging studies provide an adequate explanation of morphological deficits observed in neuropsychological populations. Considerable evidence has been amassed in the psycholinguistic literature to suggest that complex words are automatically decomposed into their constituent morphemes during non-morphological tasks such as lexical decision (Marslen-Wilson et al., 1994; Rastle, Davis, Tyler, & Marslen-Wilson, 2000). It is therefore of interest to ask whether there is a unique neural signature of the processing of morphologically complex words during tasks that do not invoke overt morphological operations.

In this study we investigate the neural correlates of the semantic processing of single words that vary in morphological complexity and syntactic class. We used fMRI to investigate the effect of these two factors and their interaction on the neural processes engaged when reading single, English words in a synonym monitoring task. This task requires equivalent and highly demanding semantic processing for all words irrespective of morphological complexity or word class, but without contamination from overt behavioural responses on critical trials (as in lexical decision) or specifically morphological operations (as in generation tasks).