G. Sartori et al. /Neuroscience Letters 390 (2005) 139-144
143
incongruent high relevance descriptions as compared with
congruent high relevance descriptions. There was no dif-
ference between congruent and incongruent low relevance
descriptions. Finally, the interactions between Category and
Relevance (F(1,23) = 0.30; p = 0.59) and between Feature
Type and Relevance (F(1,23) = 2.11; p = 0.16) were not sig-
nificant.
A similar analysis was conducted using Category (Liv-
ing versus Non-living), Relevance (High versus Low), Fea-
ture Type (Sensory versus Non-sensory), Congruency (Yes
versus No) and Laterality (CP3, CP1, CPZ, CP2, CP4)
as within-subjects factors. The absence of any Category
effect (F(1,23) = 1.79; p = 0.194), a strong Relevance effect
(F(1,23) = 33.16; p < 0.001) and also a strong Congruency
effect (F(1,23) = 28.34; p < 0.001) were confirmed. Further-
more the significant interaction between Congruency and
Laterality (F(4,92) = 4.79; p < 0.001) indexes a larger N400
on the right hemisphere sites, a result that was reported before
many times [13].
The N400 amplitude to Sensory descriptions did not dif-
fer from that of Non-sensory descriptions (F(1,23) = 0.47;
p = 0.51). This result clearly indicates that when semantic
relevance is matched among feature types any previously
reported difference in ERPs disappears [3].
Previous electrophysiological investigations using the
N400 indicated both a Category effect with larger N400 for
Living (as compared to Non-living [8,11,23]) and a feature
type effect with larger N400 for Sensory semantic features (as
compared to Non-sensory [3]). This pattern of results leaded
to contrasting interpretations. On one side, different ERPs
between categories seemed to parallel behavioural dissocia-
tions between Living and Non-living. This was interpreted as
supporting the view that categories were organising princi-
ples at neural level [2]. On the other side, the different ERPs
between feature types (Sensory versus Non-sensory) was also
considered as evidence for an organising principle based on
featural content (e.g. [14]) (Figs. 1-3).
Our data show that these may be spurious results due to
the lack of control over a parameter of semantic features that
greatly affects concept retrieval: semantic relevance. In fact,
given that lower semantic relevance is characteristic of Living
and of Sensory features [19], and given that lower relevance
Fig. 3. ERPs to High Relevance as compared to Low Relevance concept descriptions. Negativity is larger to Low Relevance descriptions.