Psychological Research (2009) 73:633–643
DOI 10.1007/s00426-008-0175-1ORIGINAL ARTICLE
Cognitive and neuronal processes involved in sequential
generation of general and speciWc mental images
Simona Gardini · Cesare Cornoldi · Rossana De Beni ·
Annalena Venneri
Received: 26 February 2008 / Accepted: 2 October 2008 / Published online: 6 November 2008
© Springer-Verlag 2008
Abstract Mental image generation is a complex process
mediated by dynamically interrelated components, e.g.
image generation and image enrichment of details. This
study investigated the cognitive and neural correlates of
sequential image generation. An event-related fMRI exper-
iment was carried out in which general and speciWc images
had to be generated sequentially in two diVerent positions.
Participants had to generate either a general image Wrst and
then a speciWc one or a speciWc image Wrst and then a gen-
eral one, in response to the same word-stimulus. Generation
times showed that speciWc images took shorter to be pro-
duced if they had been preceded by the generation of a gen-
eral image. The fMRI results showed that position of
generation and type of image was associated with diVerent
patterns of neurofunctional change. When an image was
generated as Wrst, areas of activation were found in the
parahippocampal, fusiform and occipital regions. These are
areas associated with memory retrieval and visual process-
ing. When an image was generated as second, signiWcant
activations were found in superior temporal and precuneus
areas, brain structures that are involved in the storage of
visual memory for object shapes and imagery, respectively.
The generation of a general image was supported by frontal
areas and by the precuneus. The generation of a speciWc
image involved frontal and thalamic areas (structures
associated with visual processing of details) and the posterior
cingulate cortex. When shifting from a speciWc image to a
general one, a higher level of activity was found in the middle
frontal gyrus involved in global visuo-spatial processing,
suggesting that the generation of speciWc images required
the retrieval of an object’s global shape. Altogether, these
data suggest that the sequential generation of diVerent types
of image is associated with discrete processes but also
shares common cognitive and neural components.
Introduction
Several studies have examined the cognitive properties of
mental images generated from long-term memory in
response to concrete nouns, such as generation time (Coc-
ude & Denis, 1988; Cornoldi, De Beni & Pra Baldi, 1989;
De Beni & Pazzaglia, 1995; Cocude, Charlot & Denis,
1997; De Beni et al., 2006), maintenance time (Cocude &
Denis, 1988; Cocude et al., 1997; De Beni et al., 2006),
vividness scores (De Beni & Pazzaglia, 1995) and number
of transformations (De Beni et al., 2006). The mental gen-
eration of an image is a rapid event and the visual represen-
tation of a concrete concept is considered an immediate and
spontaneous process. Generating mental images for words
facilitates their recall in a successive memory task more
eVectively than performing any verbal and/or abstract task
related to the words’ meaning (Paivio, 2006). In addition to
the cognitive processes involved in image generation, a
number of studies have also investigated the neuronal cor-
relates of imagery (e.g. Mellet, Tzourio, Crivello, Joliot,
Denis, Mazoyer, 1996; D’Esposito et al., 1997; Gardini
et al., 2005, 2006). In most instances, to have a clear and
well-deWned object of study, the process of mental genera-
tion of images has been considered as a unique and stable
S. Gardini
Department of Neuroscience, University of Parma, Parma, Italy
S. Gardini · A. Venneri
Clinical Neuroscience Centre, University of Hull, Hull, UK
C. Cornoldi (&) · R. De Beni
Department of General Psychology, University of Padua,
Via Venezia, 8, 35131 Padua, Italy
e-mail: cesare.cornoldi@unipd.it123

634 Psychological Research (2009) 73:633–643event. In so doing, these studies have neglected the fact that
imagery is a complex multi-componential process (Koss-
lyn, 1994) and that, once generated, mental images become
part of the stream of consciousness, continuously evolve
and are subject to change and decay (James, 1890; Cocude
& Denis, 1988).
When generating a mental image of an object from long-
term memory, diVerent types of image can be produced.
Cornoldi et al. (1989) and De Beni & Pazzaglia (1995) pro-
posed a taxonomy of mental images. A clear distinction
was made between general and speciWc images. A general
image was deWned as a skeletal, basic and prototypical rep-
resentation of an object, e.g. a general image of a car corre-
sponds to the global image of a car, with its basic
components, such as wheels, body, steering wheel, etc.,
without any particular details. A speciWc image was deWned
as the visual representation of a particular exemplar of an
object with its speciWc details, e.g. a speciWc image of a car
can be that of a Ferrari Testa Rossa, which is red, low,
large, with big wheels, etc. SpeciWc images refer, therefore,
to knowledge of speciWc objects, which is stored in seman-
tic memory, and are distinct from autobiographical images,
which are associated with events stored in autobiographical
memory. There are data on generation times and recall
(Cornoldi et al., 1989) supporting this distinction, which is
also reinforced by the identiWcation of diVerent neurologi-
cal correlates (Gardini, Cornoldi, De Beni & Venneri,
2006).
General and speciWc images are associated with diVerent
levels of cognitive performance (Cornoldi et al., 1989;
Helstrup et al., 1997; De Beni & Pazzaglia, 1995). General
images take less time to generate, suggesting that the typi-
cal image generation process starts with a general image.
General images are most likely the type of image produced
by participants who have not been speciWcally instructed to
generate an image of a particular type (Cornoldi et al.,
1989). SpeciWc images are those receiving the highest rat-
ings on a vividness scale and take longer to generate than
general ones (Cornoldi et al., 1989; De Beni & Pazzaglia,
1995).
The generation of a visual image from long-term mem-
ory requires the retrieval of stored information used for the
formation of a visuo-spatial pattern which is then temporar-
ily maintained into a short-term memory system, deWned as
a ‘visual buVer’ (Kosslyn, 1994) or, more simply, as the
visuo-spatial working memory system (Cornoldi & Vecchi,
2003). The generation of a mental image has been deWned
as a multi-componential process in which the image
retrieved from long-term memory is inspected at Wrst and
single parts and individual details are subsequently added.
The process has also been described as a sequential set of
cognitive operations in which prototypical semantic and
conWgural knowledge of the object is retrieved Wrst and
visual details are subsequently added to enrich that image
(Cornoldi et al., 1989; Kosslyn, 1994). According to Koss-
lyn (1994), once an image is in the short-term memory
visual buVer, an ‘attentional window’ system would take
control of those mechanisms responsible for selection and
aggregation of visual details to a skeletal image. The newly
created detailed mental representation of an object does not
correspond to a simple re-activated visual memory of that
object, but to a new and original conWguration.
Functional neuroimaging studies have oVered additional
evidence in favour of a distinction between general and
speciWc types of image. DiVerences in the pattern of brain
activation resulting from the generation of various types of
image have been observed (D’Esposito et al., 1997; Ishai,
Ungerleider & Haxby, 2000; Gardini et al., 2006). Func-
tional magnetic resonance imaging (fMRI) studies of the
generation of mental images for concrete words showed
signiWcant activation in the left fusiform gyrus, premotor
area and anterior cingulate area (D’Esposito et al., 1997).
Distinct patterns of brain activation were also seen during
the generation of general and speciWc images in response to
concrete nouns (Gardini, De Beni, Cornoldi, Bromiley &
Venneri, 2005). The generation of general images activated
a set of bilaterally distributed regions involving the frontal,
parietal and occipital areas (including the associative visual
areas). The generation of speciWc images activated the fron-
tal cortex and the medial dorsal nucleus of the thalamus.
This pattern indicates that the generation of general mental
images relies not only on the retrieval of conceptual knowl-
edge via the activity of the left temporo-frontal areas, but
might also be supported by the retrieval of deWnite visual
features through areas involved in processing visuo-spatial
information and in the formation of Gestalt-like shapes
(Lee, Mumford, Romero & Lamme, 1998). In contrast, the
generation of speciWc mental images is mediated by those
brain regions which are responsible for attention to visual
details, including the thalamus (LaBerge, 2000). These
data, therefore, corroborate the Wndings obtained from
behavioural research, which have suggested that the gener-
ation of diVerent types of image relies on distinct cognitive
mechanisms.
Despite the importance of the relationship between
diVerent types of mental images and the evidence that men-
tal imagery is dynamic in nature, the interplay between
diVerent types of mental image during the process of image
generation and maintenance has not been investigated.
There is some behavioural evidence that a certain type of
image cannot be maintained in the mind eye for long with-
out being naturally subjected to transformation (Cocude
et al., 1997; De Beni, Pazzaglia & Gardini, 2006). To our
knowledge, no study has looked at which mental operations
and related brain structures support shifting from one type
of image to another, nor has any behavioural experiment123