Action comprehension in aphasia: linguistic and non-linguistic deficits and their lesion correlates

Abstract

We tested aphasic patients’ comprehension of actions to examine processing deficits in the linguistic and non-linguistic domains and their lesion correlates. Twenty-nine left-hemisphere injured patients and 18 age-matched control subjects matched pictured actions (with the objects missing) or their linguistic equivalents (printed sentences with the object missing) to one of two visually-presented pictures of objects. Aphasic patients performed poorly not only in the linguistic domain but also in the non-linguistic domain. A subset of the patients, largely consisting of severe and non-fluent aphasics, showed a greater deficit in the linguistic domain compared with the non-linguistic domain and across the patient group, deficits in the linguistic and non-linguistic domains were not tightly correlated. Poor performance in pantomime interpretation was associated with lesions in the inferior frontal, premotor and motor cortex, a portion of somatosensory cortex, and the caudate, while poor reading comprehension of actions was associated with lesions around the anterior superior temporal lobe, the anterior insula and the anterior portion of the inferior parietal lobe. Lesion size did not correlate with deficits. The lesion results for pantomime interpretation deficits demonstrate that lesions in the frontal component of the human analog of the “mirror neuron system” are associated with deficits in non-linguistic action understanding. For reading comprehension deficits, the lesion correlates are brain areas known to be involved in linguistic tasks including sentence processing and speech articulation; the parietal lesion site may also correspond to a subpart of the human mirror neuron system. These results indicate that brain areas important for the production of language and action are also recruited in their comprehension. Similar findings have been reported in electrophysiological and neuroimaging studies. Our findings now also lend neuropsychological support to an embodied view of brain organization for action processing.

Introduction

While aphasia is primarily characterized by disturbance of language functions following brain injury, patients have been observed to also exhibit impairments in nonverbal domains, revealed by tasks such as associating pictures with corresponding objects (De Renzi, Pieczuro, & Vignolo, 1968), colors with pictures (De Renzi, Faglioni, Scotti, & Spinnler, 1972), and environmental sounds with pictures (e.g., Saygin et al., 2003a, Spinnler and Vignolo, 1966). In particular, aphasic patients’ deficits in using and recognizing signs, gestures and pantomime have been examined in numerous studies (e.g., Bell, 1994; Duffy & Duffy, 1981; Gainotti & Lemmo, 1976; Goodglass & Kaplan, 1963; Pickett, 1974; Varney, 1978; Wang & Goodglass, 1992).

In the present study, we examined aphasic patients’ comprehension of visually presented action stimuli in both linguistic and non-linguistic domains. We used a variant of a classical neuropsychological paradigm commonly used to test comprehension in aphasic patients: an object selection task. We had a two-alternative forced-choice (2AFC) design and asked patients to choose the object that best matched visually presented stimuli containing action information. In both the linguistic and the non-linguistic domains, the associated objects upon which the actions should be carried out were removed from the stimuli; thus subjects matched either a sentence missing its object (such as “he is licking the …”), or a picture missing its object (such as a picture of a boy licking an invisible ice-cream cone) to the corresponding object (in this case, the ice-cream cone). We used static black and white drawings of pantomimed actions in our design (like Seron, van der Kaa, Remitz, & van der Linden, 1979) because they are more appropriate visual matches to written text stimuli which are also static in nature.

We had two main goals in this study: (1) to test linguistic and non-linguistic action comprehension at the same time, using the same task, on the same patients, and with stimuli as closely matched as possible, (2) to conduct lesion-symptom mapping analyses using voxel-based lesion symptom mapping (VLSM; Bates et al., 2003b) to identify the lesion correlates of action comprehension deficits.

Regarding the first goal, relationships between linguistic and non-linguistic deficits in aphasic patients are important to examine because they shed light on whether aphasia is a domain-specific disorder which affects only language, or is part of a larger deficit which affects other domains as well. Such questions have been asked since the early days of neurology. Finkelnburg (1870) was the first to propose what is known as the “asymbolia” hypothesis: he suggested that a single underlying factor was common to both the language impairments in aphasia and the deficits in nonverbal domains that these patients exhibit. This idea received some support from subsequent pioneers in neurology as well (e.g., Goldstein, 1948, Head, 1926). On the other hand, it does not seem plausible that aphasia is completely reducible to a strong version of asymbolia, especially given that dissociations in performance between linguistic and non-linguistic domains can be encountered in individual patients.

Even though nonverbal deficits in aphasia have been of interest to researchers for a long time, it has been difficult to assess whether the linguistic and the non-linguistic deficits patients exhibit are related to each other. First, performance on language processing and on non-linguistic tasks must be explored in the same patients. Furthermore, task and stimulus-level factors should be as closely matched as possible. Considerations such as these motivate the first goal of the present study: to contrast linguistic and non-linguistic comprehension of action information in aphasia by comparing performance in the two domains in the same patients more directly than in previous studies. Previous studies seeking correlations between patients’ performance in various language tests and various action comprehension tests do exist, although stimuli and tasks have often not been closely matched across the two domains. While some of these studies found correlations between language impairments and non-linguistic action processing impairments in aphasic patients (e.g., Duffy & Duffy, 1981; Pickett, 1974; Seron et al., 1979, Varney, 1978, Varney, 1982), others found largely uncorrelated performance (e.g., Bell, 1994; Goodglass & Kaplan, 1963; Kimura, 1977).

With regards to the second goal, although it has been known since the early days of neurology that left-hemisphere lesions can often cause receptive and/or expressive disorders in both language and action domains (i.e., aphasic and apraxic disorders) and that patients with right hemisphere injury will rarely exhibit such impairments, the precise lesion sites leading to aphasic and apraxic deficits remain quite unclear. Specifically, results on lesion correlates of impairments in action, pantomime and gesture comprehension deficits are few, and not entirely consistent. Heilman and colleagues have reported that apraxic patients with posterior lesions have more trouble in comprehending the meaning of pantomimes (Heilman et al., 1982, Rothi et al., 1985) and have suggested that posterior parietal regions of the cortex may mediate the production and comprehension of purposeful movements (see also De Renzi et al., 1986, Kertesz et al., 1984). On the other hand, Ferro, Martins, Mariano, and Castro Caldas (1983) reported that while gesture recognition impairments were most commonly associated with parietal lesions in chronic stages of brain damage, in acute stages it was the patients with left frontal and basal ganglia damage who showed deficiencies, but unfortunately this study had a rather small sample size. Other studies failed to find reliable lesion sites associated with deficits (e.g., Schnider, Hanlon, Alexander, & Benson, 1997; Wang & Goodglass, 1992). Recently, Tranel, Kemmerer, Adolphs, Damasio, and Damasio (2003) used more novel lesion-mapping methods and reported that lesions in the left premotor/prefrontal and parietal cortex and in the white matter underlying the left posterior middle temporal cortex are implicated in deficits in tasks which were designed to tap into conceptual knowledge for actions.

There is also a substantial literature on the related question of brain areas differentially involved in the naming of actions versus objects, and/or the processing of verbs versus nouns. Many researchers have argued that left frontal areas are differentially involved in the processing of actions or verbs. For example, in a PET study using a lexical decision task, Perani, Cappa, Schnur, and Tettamanti (1999) found that verbs activated left dorsolateral frontal cortex more than nouns. However, other studies have failed to find significant differences; for instance, Tyler, Russell, Fadili, and Moss (2001), using carefully matched stimuli, did not find any regions differentially involved in the lexical decision or semantic processing of nouns versus verbs. Hillis, Tuffiash, Wityk, and Barker (2002) reported that damage or hypoperfusion in precentral and middle temporal gyri were associated with action naming deficits in patients with acute left hemisphere injury, while for object naming, middle temporal and superior temporal gyri were associated with impairment. However, for comprehension of action and object words, they did not find separate sites; impairments were associated with superior temporal lesions. Hillis et al. suggested in light of this finding that only the naming of actions, rather than semantic knowledge, may be localized to left frontal cortex.

Given the diverse results which have been reported in the literature, we wanted to use VLSM, a quantitative lesion-symptom mapping technique, to contribute to identifying lesion correlates of action comprehension in aphasia in linguistic and non-linguistic domains.

In addition to these two major goals, we had some other points in mind in our design: In line with earlier studies (e.g., Varney, 1978), we also addressed the effect of semantic competition in both domains in order to see if processing in the two domains is similarly modulated by higher-level conceptual constraints. In addition, following Seron et al. (1979), Wang and Goodglass (1992) and Bell (1994), we also used distracters that were related to the targets in the way they may be handled, to see if there are differential effects of this kind of competition (previous researchers termed these morphological, perceptual or motoric distracters; here we refer to these as “affordance-based” distracters; see Gibson, 1977).