In this issue of JINS, Gasparini and her colleagues (Reference Gasparini, Hufty, Masciarelli, Ottaviani, Angeloni, Lenzi and Bruno2008) report a subject with a right anterior choroidal artery territory stroke that damaged the anterior hippocampus, but likely more importantly, much of the ventral posterior limb of the internal capsule, effectively disconnecting the right thalamus from temporal cortex. Through systematic cognitive neuropsychological detective work, they provide a compelling case that the essential deficit was one of working memory. This is potentially a very important discovery. The thalamus is a phylogenetically ancient structure, and there is considerable evidence that, in human beings, much of its function has been subsumed by cortical mechanisms (Nadeau & Crosson, Reference Nadeau and Crosson1997). Consequently, there are fundamental limits to how much animal studies can inform us about human thalamic functions, and we are particularly reliant on cognitive neuropsychological studies like this one. The work of Gasparini et al. also provides a fine demonstration of the continuing value of cognitive neuropsychology in advancing our understanding of the details of brain function.
Working memory corresponds to the temporary recruitment of selected neural networks needed for task performance. The term “working memory” is a misnomer, because memory refers to stored information, whereas “working memory” refers to information that is temporarily brought out of the mental library for viewing. Other terms often used interchangeably, equally misleading, include “activation” and “attention.” Working memory corresponds neurophysiologically to several different processes, including alterations in neural firing rates, alterations in likelihood of neural firing, and selection of inputs that will produce neuronal firing in engaged networks (Desimone & Duncan, Reference Desimone and Duncan1995; Moran & Desimone, Reference Moran and Desimone1985). Several years ago, we coined the less encumbered term “selective engagement” to refer to these processes (Nadeau & Crosson, Reference Nadeau and Crosson1997). Most studies of working memory in human subjects have focused on the cortex, particularly the prefrontal cortex. Over 30 years ago, Chavis and Pandya (Reference Chavis and Pandya1976) published a landmark paper in which they demonstrated exquisitely mapped connections between the full extent of frontal cortices and posterior sensory, polymodal and supramodal association cortices. Subsequent studies have confirmed their findings and shown these connections to be reciprocal (Constantinidis & Procyk, Reference Constantinidis and Procyk2004). This connectivity provides the basis for two major phenomena. Prefrontal to posterior projections provide the basis for selective engagement of particular posterior cortical networks, that is, working memory. Reciprocal posterior to frontal connections serve to fully inform the frontal cortex of what is going on posteriorly—the likely basis of the “mirror neuron” phenomenon (Rizzolatti & Craighero, Reference Rizzolatti and Craighero2004). In essence, the result is that whatever we perceive has immediate implications for frontally mediated plan formulation and selection. It, furthermore, seems likely that the entire cerebrum is involved in working memory and mirror neuron phenomena, broadly conceived, and that our current notions that these functions are more localized reflect the limitations of our experimental paradigms rather than the limitations of the brain substrate for these phenomena.
There is good evidence that selective engagement is also subserved by subcortical mechanisms, including the thalamus and biogenic amine systems (Nadeau, Reference Nadeau and Noseworthy2006; Nadeau & Crosson, Reference Nadeau and Crosson1997). We know that the thalamus functions as a gated relay, most obviously gating the transmission of visual, auditory and sensory information through the lateral and medial geniculate nuclei and the ventral posterior nucleus, respectively, to the cortex. However, it appears that these elemental sensory gating functions have largely been superceded in human beings by cortically based gating processes subserving attention, and that in human subjects, thalamic gating predominantly affects cortico-thalamo-cortical transmission. Gating though the thalamus is regulated in a more or less all or none manner via input from brainstem structures, most notably the midbrain reticular formation and noradrenergic and serotonergic projections. Selective gating by the thalamus is regulated in two ways: (1) via direct projections from overlying cortices and (2) via projections from prefrontal cortex, involving the thalamic reticular and intralaminar nuclei (central median and parafascicular). This frontal system, scarcely studied in animals (Bailey & Mair, Reference Bailey and Mair2005; Minamimoto & Kimura, Reference Minamimoto and Kimura2002; Weese et al., Reference Weese, Phillips and Brown1999), is of particular interest because it is likely to be involved in thalamically mediated volitional selective engagement processes, such as those involved in language production and working memory.
Left thalamic lesions cause an aphasia marked by particular difficulty in translating semantic representations into spoken and written word representations (suggesting a problem with semantic access), but in nearly all cases, semantic knowledge, phonological sequence knowledge, and grammatic function are preserved (Nadeau & Crosson, Reference Nadeau and Crosson1997; Nadeau & Rothi, Reference Nadeau, Rothi and Chapey2008). Raymer et al. (Reference Raymer, Moberg, Crosson, Nadeau and Gonzalez Rothi1997) argued that translating semantic into word representations makes particularly great demands on semantic engagement because there is one and only one correct response, requiring extraordinary precision of engagement. Certain left thalamic lesions can blunt this precision, leading to visual/semantic paraphasic errors and nonresponses even as semantic function is preserved. The discovery by Gasparini et al. of deficits in visual imagery and image completion from individual features or partial representations, coupled with spared perceptual processes, visual knowledge, and recognition abilities, may be evidence of an analogous loss of precision of volitional selective engagement in the right hemisphere, or impairment in a particular dimension of selective engagement mediated by the thalamus, affecting flexibility of selective engagement.
Gasparini et al. rightly point out that the hippocampal infarction could be a contributor to the deficits they found. The hippocampus supports fast Hebbian learning, encoding new information over a time span of roughly a second (the short phase of long-term potentiation). Thus, there is an overwhelming likelihood that the hippocampus contributes to many phenomena that have been subsumed under the concept of working memory, particularly when longer delays are involved (Shrager et al., Reference Shrager, Levy, Hopkins and Squire2008). For example, as Gasparini et al. point out, primacy effects in supraspan memory tasks disappear in the presence of hippocampal lesions. Thus, there is a need to replicate their findings in subjects with thalamic lesions, for example in the tuberothalamic or paramedian artery territories, which produce a pure thalamic lesion with no hippocampal involvement.
