difference was larger than 6 of 8 patients who showed a
significant difference using the Revised Standardized
Difference Test [RSDT]). The absolute size of the differ-
ence is therefore not a definitive marker for the pres-
ence or absence of a differential deficit. Moreover,
consider patients GE (90 vs. 86%) and ZG (90 vs.
82%), who showed better absolute feature verification
for living than for nonliving things, yet had differential
living deficits (indeed, classical dissociations). So, not
only do absolute differences provide no guarantee to the
presence of a differential deficit, they provide no guar-
antee about the direction of any deficit.
GENERAL DISCUSSION
This study investigated the validity of the criteria that are
commonly used to document the presence of category-
specific deficits, in particular, the strong dependence
upon within-patient comparisons of absolute living and
nonliving scores rather than comparing patient perform-
ance with that of healthy controls. Additionally, most
studies have relied upon establishing category disorders
on picture-naming tasks; and in some studies (especially
those examining Alzheimer’s patients), rely solely upon
picture naming as evidence of category specificity (Laws,
Gale, et al., 2005). The current study illustrates how such
methods lead to erroneous conclusions about both the
presence and even the type of category deficit exhib-
ited by patients. These findings have implications for a
range of questions relating to category disorders and
raise methodological and theoretical issues regarding
what might be regarded as acceptable evidence for
dissociations and double dissociations—especially when
used to make claims about the fractionation of cogni-
tive architecture.
Use of Chi-square Analysis
The current study highlights the unreliability of using
within-patient comparison of absolute scores. As already
mentioned, studies of Alzheimer’s patients frequently
assess picture naming only, and in line with the data
presented here, would be likely to misrepresent not
only the incidence, but even the direction of a category
effect. The nature of x2 analyses means, of course, that
studies always assume the deficit to be ref lected in the
smaller of any two scores obtained. The current study
shows how this metric is unreliable and confirms our
recent findings in two other groups of Alzheimer’s pa-
tients (Laws, Gale, et al., 2005).
The widespread use of within-patient comparisons to
determine category deficits may well have distorted the
incidence, and therefore, the ratio of living to nonliving
deficits reported in the literature. Even if patients show a
patently large absolute difference across category, this
does not necessarily denote a category disorder; or
conversely, does the failure to find a large absolute
difference across category refute the presence of a cat-
egory deficit. Indeed, several HSE and Alzheimer’s pa-
tients showed large absolute differences that were sig-
nificant using within-patient x2 analyses, but not when
referenced to control data (e.g., HSE patients such as
BM, DD, and SG for picture naming; and Alzheimer’s
patients such as RN, MR, and VP for naming-to-
description). Conversely and more commonly, x2 analy-
ses led to false negatives in both HSE (11 dissociations)
and Alzheimer’s (7 dissociations) patients. This ref lects
the simple fact that absolute differences are critical for
demonstrating deficits when using x2; however, healthy
controls may show category advantages, and therefore,
even a lack of absolute difference in patients may be
important. For example, analysis of picture-naming for
HSV patients DD and SG indicated no category differ-
ence (56 vs. 57% and 76 vs. 78%), yet both showed sig-
nificant dissociations when referenced to control data.
Conversely, Alzheimer’s patient FR showed a 31% dif-
ference in living and nonliving naming-to-description
(19 vs. 50%), but this was not significant when referenced
to control data.
Picture Naming as Indicator
of Semantic Impairment
As outlined in the Introduction, the extant literature
relies heavily upon using impaired picture naming to
identify category effects; other tests may be used subse-
quently to elaborate on this, but picture naming has
assumed a role as the sine qua non test of category
specificity. Therefore, the failure to find a category
difference on naming tasks is typically viewed as under-
mining the presence of any category effect (Laws, 1998;
Laws, Evans, et al., 1995).
Although most cognitive psychologists are comfort-
able with the notion of impaired naming and intact
semantics (i.e., lexical deficits), the converse—impaired
semantics and intact naming—have been less frequently
reported, perhaps because of the assumption that it
ought not to happen within most currently accepted
models (Laws, Evans, et al., 1995). Nonetheless, the
current study documents several Alzheimer’s (2/15:
13%)1 and HSE (3/15: 20%) patients who showed a
category effect in semantics (using feature verification),
but no category effect on either of two naming tasks.
Some patients (2/19 Alzheimer’s and 4/15 HSE)
showed category deficits on tasks other than picture
naming, including patients (e.g., Alzheimer’s patient FB
and 2 HSE patients DD and TL) with impaired perform-
ance on all tasks except picture naming. Therefore,
using picture naming as a critical test of category effects
would miss some patients who showed consistent def-
icits on other tasks. Indeed, reliance on any one task
may lead to a quite different conclusion from that if
another task had been chosen. Moreover, feature veri-
1456 Journal of Cognitive Neuroscience
Volume 17, Number 9