A Gradient in Retrograde Amnesia

In the NVT, neologisms are ordered into decades by the year that they entered the language. These years are only indicative of when memories for the word could have started to be formed: the date that a word enters a language is not the date that most people first experience it. Moreover, words usually do not leave the language, so that memories for the word can continuously have formed since the moment the word entered the language to the onset of amnesia. Nevertheless, Verfaellie et al. (1995) reported a gradient in the remote memory of patients with Korsakoff's disease using a neologisms test. Gradients can thus be observed with such retrograde amnesia tests. Presumably, a memory deficit that disproportionately affects recent memories would affect neologisms that have been encountered only recently, while sparing those that were also stored in remote time periods.

Here, no gradient could be established within the NVT, though one was found in episodic retrograde amnesia. The inconsistent findings are in line with the literature, as gradients are sometimes found in patients with Alzheimer's disease (Beatty & Salmon, 1991; Beatty et al., 1988; Brown, 2002; Kopelman et al., 1989; Moscovitch, 1982), and sometimes not (Leplow et al., 1997; Thompson et al., 2002; Wilson et al., 1981). It should be noted that the periods tested in the NVT all fall between the ‘recent’ period on the AMI and the other two periods. The time range of the NVT may thus have been too small to find a gradient.

Semantic Memory and Alzheimer's Disease

The present findings also highlight the importance of semantic memory disturbances in Alzheimer's disease. This is consistent with previous findings with other semantic memory tasks (Petersen et al., 1999; Salmon et al., 1992). The finding that patients with Alzheimer's disease do not benefit as much from cueing as normal controls (Cahn et al., 1995) has also been interpreted as indicative of semantic memory problems.

An interesting finding is the strong correlation between the NVT and the NART. Recent evidence suggests that NART scores decline as dementia progresses, and that it may thus underestimate premorbid IQ in patients with more severe dementia (Cockburn et al., 2000; Schmand et al., 1998). The decline of the NART-based estimate of IQ was related to deterioration of semantic memory as reflected in verbal abstraction and category fluency (Schmand et al., 1998). The current result corroborates that finding, suggesting that the NART is indeed sensitive to declines in semantic memory. It does not exclude an alternative explanation, however, namely that semantic memory tests are sensitive to crystallized intelligence as purportedly measured by the NART (the fact that the NVT and NART also correlate within the normal control group supports this alternative). Indeed, tests of retrograde amnesia typically correlate with intelligence (Kapur et al., 1998).

Semantic Retrograde Amnesia

In many ways, it is not surprising that retrograde amnesia should extend to semantic memory. Nearly all tests of semantic memory are tests of remote memory in the sense that they measure old memories. Moreover, all remote memory tests contain semantic items, at least following some definitions of semantic memory. If semantic memory is “the component of long term memory which represents our knowledge of objects, facts and concepts, as well as words and their meanings” (Garrard et al., 1997), then all tests of public knowledge are tests of semantic memory (as they query facts about the news), and many tests of autobiographical memory contain semantic items (the so-called “personal-semantic” questions, Kazui et al., 2003; Kopelman et al., 1989). Nevertheless, some semantic memories are more semantic than others. Semantic memories are “usually overlearned and not temporally specific” (Garrard et al., 1997). Knowledge of news events need not be overlearned, and is to some extent temporally specific. Although some “personal-semantic” questions in the AMI lack this specificity (such as items in which subjects must generate addresses on which they lived), others are specific about time (such as the many questions about a specific wedding in the “young adult” section of the AMI, Kopelman et al., 1989). Lexical knowledge is the prototype of semantic memory; tests that use word meanings as its material, such as the NVT, are therefore unequivocally semantic. The current results thus seem to point to the existence of semantic retrograde amnesia.

Semantic retrograde amnesia is only one explanation for the present results. Three alternative accounts can be envisioned. Since AD has an insidious onset the deficits may not reflect retrograde, but anterograde amnesia. To explain why AD patients also show deficits on the most remote neologisms, one might claim that words need constant repetition in order to remain in lexical memory. Thus recent anterograde amnesia might produce a loss of words learned a long time before. This explanation seems unlikely, however, as several studies have suggested that knowledge enters a state of permastore after some four years (Bahrick, 1984, 1992; Conway et al., 1991).

An alternative suggestion would be that word definition is a task relying partly on episodic memory. Although at first glance this seems unlikely, a similar hypothesis has recently been leveled for the quintessentially semantic task of category fluency: category members may often be generated with help of episodic memories (Hayes et al., 2001).

A third explanation would question whether there is a neuropsychological dissociation in remote memory between episodic and semantic memories. In a recent review, Squire and Zola (1998) argued that the case for such a dissociation is not as strong as is commonly thought. Episodic amnesia combined with intact semantic knowledge may in most cases be a dissociation between a learning deficit leading to an inability to form new episodic memories, and intact old memories. Although patients with normal semantic learning in the face of episodic amnesia have also been reported (e.g., Vargha-Khadem et al., 1997), Squire and Zola (1998) argue that residual episodic memory may explain the acquisition of semantic memories in these patients. Patients with semantic memory deficits and intact episodic memory are at most very rare. It was for a time thought that patients with semantic dementia presented with such a pattern (Graham & Hodges, 1997; Hodges et al., 1992; Snowden et al., 1996), but recently it has become clear that episodic memory is by no means intact in semantic dementia (Graham et al., 2000; Westmacott et al., 2001).

It may be that extended retrograde amnesia is in fact a semantic memory deficit—this is the portent of one view on temporally graded retrograde amnesia, semantization (Cermak, 1984; Meeter & Murre, 2004; Rosenbaum et al., 2001). In this view, episodic memories become semantized with time, which is equivalent to losing their temporal and contextual specificity. This would predict that extensive retrograde amnesia could only occur in the presence of semantic memory deficits, and that loss of semantic and remote episodic memories would stretch back in time to the same extent. The current results do not speak to this conjecture, as the episodic retrograde amnesia test used had a different time scale than the NVT.