Memory, 2013

Vol. 21, No. 2, 261
279, http://dx.doi.org/10.1080/09658211.2012.725738

Object specicity and personal relevance in

long-term visual remembering
Peter Walker1, Louise Mahon1, Helen Kennedy1, and Damon Berridge2
1

Department of Psychology, Lancaster University, Lancaster, UK

Department of Mathematics and Statistics, Lancaster University, Lancaster, UK

The personal relevance of an object is multi-faceted, each facet being capable of contributing to the
effects on object memory attributed to personal relevance. An objects status as an individual object
(object specificity), rather than just a category of object, is one such facet and its impact on the long-term
visual remembering of everyday objects is assessed in two experiments. Images and drawings were
produced under generic (e.g., Please draw a bed) and personal exemplar (e.g., Please draw your
bed) instructions, and participants indicated the degree to which the image on which their drawing was
based was of a specific object or a generic object. Object specificity induced a sense of time and place for
a remembered object, the most recent encounter with the object being most salient. Other aspects of
personal relevance collectively facilitated the retrieval of an objects category-irrelevant features
(thereby increasing the vividness of the object image), the other objects with which it was seen, and
a more general episodic sense of place. Against a broader theoretical perspective, it is proposed that
visual episodic memory and visual knowledge are primary sources of information for specific personally
relevant objects and generic objects, respectively.

Keywords: Visual knowledge; Visual episodic memory; Personal relevance; Object specificity; Visual imagery;
Drawing from memory.

Two recent studies highlight the significance of

people presuming ownership of an item for how
well they later retrieve information about it from
recent memory (Cunningham, Turk, Macdonald,
& Macrae, 2008; Van den Bos, Cunningham,
Conway, & Turk, 2010). Participants were shown
pictures of individual supermarket items, with
each pictured item being accompanied by a
colour cue indicating whether it should be placed
in their shopping basket or the basket of a
fictitious other person. Later recognition memory
was superior for items of which the participants
presumed ownership.
The personal relevance of an item is considered
to be the important factor here, with ownership
being just one of several ways of ensuring that an

item takes on such relevance (Cunningham et al.,

2008; Van den Bos et al., 2010). Because imbuing
an item with personal relevance normally entails
changing several things about it, it follows that
there can be various ways in which such relevance
influences object memory. For example, personal
relevance normally makes an item more important, more valuable as something that can be
traded, and more positively valenced. For any of
these reasons, such an item could be allocated
additional cognitive resources during encoding
and retention (Beggan, 1992; Belk, 1988, 1991;
Cunningham et al., 2008; Kahneman, Knetsch &
Thaler, 1991; Knetsch & Sinden, 1984). Personal
relevance might also ensure that an item is
encoded, in part, by relating it to well-established

Address correspondence to: Peter Walker, Department of Psychology, Lancaster University, Lancaster LA1 4YF, UK. E-mail:
p.walker@lancaster.ac.uk

# 2013 Taylor & Francis

262

WALKER ET AL.

schemas concerning a persons self-concept. The

result would be a memory representation that is
elaborated to a greater extent than is the case
for items lacking personal relevance (Conway &
Dewhurst, 1995; Rogers, Kuiper, & Kirker, 1977;
Symons & Johnson, 1997). Indeed, personally
relevant items might help to shape a persons
self-concept.
Van den Bos et al. (2010) point out that
increasing the personal relevance of an object
highlights its status as a specific individual object,
over and above its status as a member of a
familiar object category. This is because personally relevant objects need to be distinguished
from other objects of the same type. This requires
the unique, category-irrelevant features identifying these specific objects (e.g., my car from
among similar cars) to be remembered. It also
encourages us to remember a range of other facts
about the objects, such as what state they are in
(e.g., whether my car is taxed and insured,
whether it needs petrol) and where they can be
found. Details about where personally relevant
objects have previously been encountered will
include the relative positions and identities of
other objects with which they were encountered.
Furthermore, given the functional significance of
the most recent of the locations in which an
individual, personally relevant item has been
encountered (because this is also likely to be its
current location), it becomes important to preserve some information about the time of each
previous encounter with it, at least sufficient to
allow the most recent location to be distinguished
from earlier locations. Of course, it would
not normally be important to retain such additional information about objects lacking personal
relevance.
It follows from this that the object specificity
induced by personal relevance should encourage
more varied information to be retained about
objects and our encounters with them. This should
not simply make the same way of remembering
the same information more efficient (although it
could do so, see below), it should also alter the
nature of the information that is remembered. In
this regard it is relevant to note that Van den Bos
et al. (2010) found that enhancing the personal
relevance of an item increased the relative involvement of episodic memory in the retrieval of
information about the item from recent memory.
Their participants reported an enhanced sense
of conscious recollection (i.e., remembering
rather than knowing) in relation to items of

which they had presumed ownership. This accords

with the need to have available information about
where individual personally relevant items can be
found (i.e., the spatio-temporal context in which
they were last encountered) and the condition
they are in. The results of Van den Bos et al.
provide strong indication, therefore, that personal
relevance, and the object specificity it entails,
change what we remember about objects, not just
how well we remember them.
The extent to which object specificity influences object memory independently of other
factors linked to personal relevance is unknown.
At one extreme it could have no influence. At the
other extreme it could explain all the effects on
memory previously attributed to personal relevance. The present study addresses this issue and
provides evidence for an intermediate position in
which object specificity and personal relevance
both make separate and distinct contributions to
object memory.
There is a relatively obvious way in which
preserving an objects unique features could improve performance in a standard recognition test
of memory. In such a test recognition foils are
normally selected from the same object categories
as the targets, making category-irrelevant features
indispensible for distinguishing targets from foils.
For example, a previously presented blue drinking
mug might be presented in a later recognition test
(as a target) alongside a green drinking mug (as
a foil) that wasnt presented previously. Clearly,
correctly accepting the target while rejecting the
foil requires information about the colour of the
to-be-remembered mug to have been retained
(with other category-irrelevant features also being
useful). Recognition memory performance will
not normally benefit from retaining information about the features shared by all mugs (i.e.,
category-defining features).
Because there is uncertainty about the level of
categorisation at which targets and foils were
matched by Cunningham et al. (2008) and Van
den Bos et al. (2010),1 the differential retention of

These two studies, as reported, leave some uncertainty

about what might be the basis for the effects of personal
relevance they observe. Some of this uncertainty arises from
the lack of information about the stimuli. Cunningham et al.
and Van den Bos et al. refer to sets of stimuli throughout.
They refer to all the stimuli they use as a set of stimuli,
and then also refer to the three equivalent sets formed from
this set. Although they go on to say that items across their

LONG-TERM VISUAL REMEMBERING

object-specific features, induced by increasing

personal relevance, could have mediated the
memory effects they observed. This would be
possible, for example, if supermarket items being
placed in a participants own shopping basket
were encoded and remembered as specific objects, embracing their category-irrelevant features
alongside their category-defining features (e.g., a
specific mobile phone, including make, model,
and colour), whereas items being placed in the
basket of a fictitious other person were encoded
and remembered only as categories of object,
embracing only category-defining features (e.g.,
just the concept of a mobile phone). In this case
the functional significance for memory of personal relevance would be mediated by the enhanced
retention of category-irrelevant features induced
by object specificity.
The two experiments reported here were
designed to determine if object specificity influences the retrieval of object information from
long-term memory, separately from the influence
of other factors linked to personal relevance.
They were also designed to reveal if object specificity influences object memory in ways consistent with the preceding discussion. For example,
will the retrieval of information about specific
objects (regardless of their personal relevance),
compared to generic objects, tend to include
more elements of the spatio-temporal context in
which the objects were encountered (i.e., will the
information be based on episodic visual memory)? Will information about the most recent
locations in which objects were encountered be
especially prominent when information is being
retrieved about specific objects? What influence, if any, will the other features linked to
personal relevance, collectively, have on object
memory?
Participants were asked to image and draw
named objects from long-term visual memory.

They were asked in two ways, one encouraging

the retrieval of information about specific objects having personal relevance (e.g., Please
draw your bed), the other encouraging the
retrieval of information about generic objects
having no such relevance (e.g., Please draw a
bed).2 In this way the focus of interest was
on the retrieval of information about objects that
had been encoded and retained in everyday
life. The shift away from examining how temporary ownership in an experimental situation
can impact on temporary (i.e., relatively shortterm) memory (see Cunningham et al., 2008;
Van den Bos et al., 2010) served to extend the
generality of previous work concerning the
influence of personal relevance on object memory. The shift also enhanced the ecological
validity of the study in relation to the conditions
under which the objects had originally been
encoded and committed to memory, there
being no attempt to control or manipulate how
objects varying in personal relevance were originally encoded. We chose to monitor aspects
of imaging and drawing partly because of a
broader research agenda concerning the interdependencies between these two products of
visual cognition.
A relatively direct approach would be to think
of personal relevance and object specificity as
dichotomous factors and create four alternative
task instructions by crossing personal relevance
(relevant versus irrelevant) with object specificity
(specific versus generic) in a 22 design (see
Figure 1A). For the relevant/generic and irrelevant/specific conditions, however, this approach
lacks feasibility because of the difficulty of
ascertaining, in advance of the study, which
generic objects (i.e., object categories) have
personal relevance for a participant, and which
specific objects lacking personal relevance
a participant will be familiar with (see panels
2

three sets were matched for item type (giving fruit and
confectionery as examples of such types), the matching
could be at this or any lower level in the taxonomy of
supermarket items. It is unclear, therefore, how and to what
extent foils were matched with targets in their recognition
tests. To illustrate, two examples of electrical item serving as a
target and a foil could be two different toasters or a toaster
and a kettle. The information needed to distinguish target
from foil is very different in these two situations (typical
names would suffice in the case of a toaster and kettle, but not
in the case of two toasters).

263

Although reference to your in the personally relevant

instructions implies ownership, it does not necessarily do so
in the strict legalistic sense of this term; that is, the sense
that allows someone to dispose of an item. The students who
drew their bed in the present study very likely did not own
it in a legalistic sense, but instead simply had primary claim
on its use. Either their parents or the university will have
owned it, depending whether they drew their bed in the family
home or their bed in university accommodation. This illustrates the sense of ownership that is made most salient by
inclusion of the term your in the personal exemplar
instructions.

264

WALKER ET AL.

i and iv in Figure 1A).3 For the relevant/specific

and irrelevant/generic conditions, however, feasible instructions can be found, albeit with some
limitations regarding the degree to which they
will be able to fix the level of personal relevance
and object specificity for a named item (see
panels ii & iii in Figure 1A). For example,
participants might be asked to Please draw
your bed in the relevant/specific condition, and
to Please draw a bed in the irrelevant/generic
condition. Provided it can be assumed that all
participants will have an exemplar from each
of the categories to be named, the first type
of instruction will help to ensure participants
images and drawings are of relatively specific
objects having personal relevance. Although the
second type of instruction leaves participants free
to draw specific objects to represent the named
categories (see Cornoldi, De Beni & Pra Baldi,
1989; De Beni & Pazzaglia, 1995, for supporting
evidence in relation to visual imagery), it does
much to encourage them to image and draw
generic objects.4
Of course, neither type of instruction will
completely constrain participants to image and
draw either a purely specific object or a totally
generic object. It is known, for example, that even
when people intend to draw specific objects, their
drawings contain a mix of object-specific and
generic elements, with the balance of this mix
varying across drawings (e.g., in drawing their
bed, a participant might reproduce the distinctive
patterning on the headboard, but not the distinctive profile of the legs, depicting these instead as
simple straight lines) (see Walker et al., 2006, and
Bozeat et al., 2003, for examples of how object
3
It is assumed that an object category is less likely to have
personal relevance than a specific object. Exceptions would
arise, for example, in relation to collectors and manufacturers
of object categories (e.g., the category of walking sticks would
be personally relevant to a collector of walking sticks) and to
people for whom a category of object has special importance
(e.g., a disabled person who needs support from a walking
stick).
4
The tendency to access a generic representation with this
type of instruction is consistent with evidence that the names
used most frequently to refer to objects are names for object
categories and, as such, have privileged links with visual
representations at an intermediate (basic) level of categorisation (Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976).
Thus visual representations at the basic level are evoked by
the presentation of a category label more frequently and more
quickly than visual representations at other levels of categorisation (Cornoldi et al., 1989; De Beni & Pazzaglia, 1995;
Gardini, De Beni & Cornoldi, 2004; Hoffmann, Denis, &
Ziessler, 1983; Rosch et al., 1976).

drawings can reflect a mix of object-specific and

generic elements). It is also known that most
visual images of named items are initially generic
in nature, and only then are elaborated to become
more specific (Cornoldi et al., 1989; De Beni &
Pazzaglia, 1995; Kosslyn, 1981). At some intermediate points in time, therefore, the images will
be a mix of object-specific and generic elements.
The implications of the presence of such a mix are
that object specificity is able to vary somewhat
independently of personal relevance, allowing the
distinct contributions of these two factors to be
identified.
Because the association between instruction
type and each of object specificity and personal
relevance is probabilistic in nature (see Figure
1B), it was acknowledged at the outset that
statistical modelling, based on regression analyses, would be needed to distinguish the influence of each factor. To allow this dissociation it
was arranged for participants to indicate how
specific versus generic was the object in each of
the images on which their drawings were based.
Instruction type (i.e., generic instructions, such as
Please draw a bed, versus personal exemplar
instructions, such as Please draw your bed) and
this index of object specificity were then to be
entered into regression analyses to see which
aspects of long-term visual remembering were
explained by object specificity, and which by all
the other factors collectively linked to personal
relevance. The steps in the argument were as
follows. It was to be determined first how much of
the variation in different aspects of the imaging
and drawing of named items was explained by
instruction type on its own. This variation was to
be seen as reflecting the impact of object specificity and/or personal relevance. If adding object
specificity to instruction type in the statistical
model then failed to add significantly to the
explanatory power of the model, the conclusion
would be that object specificity is not responsible
for any of the effects on imagery and drawing,
which instead would be attributed solely to
personal relevance. However, if adding object
specificity explained all the effects previously
attributed to instruction type, the conclusion
would be that the involvement of personal relevance is completely mediated by the object specificity it induces. Alternatively, and a priori more
likely, if a significant influence of instruction
type remained after adding object specificity to
the model, the conclusion would be that personal
relevance has an independent influence on visual

LONG-TERM VISUAL REMEMBERING

265

personally irrelevant personally relevant

Draw your bed

I

II

III

IV

specific object

generic object

Draw a bed

object specificity

personal relevance
Figure 1. (A) The different sampling spaces associated with personal exemplar and generic instructions, with personal relevance
and object specificity represented as dichotomous variables. Shading identifies those regions of the sampling space that can be made
salient by appropriate instructions. For the unshaded regions, equivalent instructions are not available. The two elliptical sampling
spaces captured by personal exemplar instructions and generic instructions illustrate how there is the opportunity for independent
variation of object specificity and personal relevance. (B) An equivalent representation of the sampling space for both types of
instruction combined, with personal relevance and object specificity both now being represented as continuous variables.

remembering, which remains after any contribution from object specificity is accounted for.

EXPERIMENT 1: IMAGING AND

DRAWING MY CHAIR VERSUS A CHAIR
Some of the types of information about familiar
objects made salient by object specificity, and by
personal relevance more generally (see above),
were monitored to identify the distinct influence
these two factors have on memory. For example,
personal relevance is expected to highlight an
objects unique (i.e., category-irrelevant) features
because, as pointed out already, it tends to be
specific individual objects that are personally
relevant. In addition it is expected that object

specificity will enhance memory for the spatiotemporal contexts in which objects were previously encountered (and, conversely, there is
little need to remember such information in
relation to generic objects).
In addition to monitoring the information
being retrieved during imaging and drawing, the
time course of imaging was monitored. Previous
research has shown that images of specific items
take longer to evoke than generic images, and
that images with episodic reference take longer
to evoke than images lacking such reference
(Cornoldi et al., 1989; De Beni & Pazzaglia,
1995; Gardini et al., 2004; Kosslyn, 1981; Palladino
& De Beni, 2003). The time to evoke each image
was to be entered into multiple regression analyses
to clarify if personal relevance and/or object

266

WALKER ET AL.

specificity influence the fluency (speed) with

which images are evoked. It is possible, for
example, that it takes longer to access details of
previous encounters with an object (i.e., access
visual episodic memory) than to access generic
visual information lacking episodic reference
(i.e., visual knowledge). In other words, anything
that increases the involvement of episodic memory will lengthen image evocation times. Preliminary exploration of these issues used analysis of
variance. This was followed up with statistical
modelling designed to reveal the underlying pattern of interdependencies among the measured
aspects of imagery and drawing. As stated already,
the central aim of the modelling was to assess the
extent to which object specificity and personal
relevance separately influence distinct aspects of
long-term visual memory when this is supporting
the imaging and drawing of named objects.

Method
Participants
The sample comprised 20 students and administrative staff at Lancaster University (10 males,
10 females) with a mean age of 25.9 years
(range19 to 58 years, SD 11.0 years). No
payment was given for their participation.

Materials
Two of the authors (LM & PW) noted the
objects that came to mind when trying to think of
instances of common items that are seen and used
by most people on most days. It was from the pool
of items generated in this way (rather than from
an established database) that 20 were selected as
the to-be-drawn items: cup, bed, toothbrush, sofa,
fridge, bag, phone, front door, cooker, lamp, shoe,
wardrobe, wallet/purse, mirror, table, television,
house, computer, shower, belt. The items were
randomly assigned to two lists of 10 items each.5
5
For both experiments reported here, the rooms in which
participants worked did not have any of the items to be drawn
on display. Therefore there were no models in the room on
which participants could image and draw the named items.
With regard to any relevant items that might be brought in by
participants themselves, the experimenter surreptitiously
checked that these were not in view during the study (e.g.,
that their shoes remained out of sight under the table on which
they worked).

Image vividness. Participants were provided

with a black fine-tip Pilot pen and 20 A4 response
sheets. On the top side of each response sheet was
a 5-point imagery vividness scale (1very unclear/hazy, 2 fairly unclear/hazy, 3clear but
not at all vivid, 4 quite vivid, 5 very vivid).
Beneath this was an empty rectangle measuring
2015 cm in which participants were to draw the
named item.
Image specificity. On the reverse side of each
response sheet was a 5-point scale concerning the
extent to which the completed drawing was based
on an image that was typical for the objects
category, rather than on an image of a particular
individual object (1 purely a generic/ prototypical item, 2 more of a generic/prototypical item
than a specific/particular item, 3 partly a generic/
prototypical item and partly a specific/particular
item, 4more of a specific/particular item than a
generic/prototypical item, 5 purely a specific/
particular item).
Episodic reference. On the reverse side of the
response sheet there was also a 5-point scale
concerning the extent to which there was episodic
reference in the participants image. This was
assessed with a single question in which episodic
reference was addressed by relating it to recollection of a particular past occasion. Thus participants were asked to use the scale to indicate the
extent to which their image was based on a
particular past occasion when they had encountered the named item (1not at all, 2 not really,
3 I imagined some aspects of a particular occasion when I encountered the imaged item, 4I
imagined most aspects of a particular occasion
when I encountered the imaged item, 5I imagined the whole of a particular occasion when
I encountered the imaged item).

Design
All participants were asked to image and draw
all 20 items. A subset of 10 items was introduced
with generic instructions (e.g., Please draw a
cup). The other subset, dealt with in a separate
block of trials, was introduced with personal
exemplar instructions (e.g., Please draw your
cup). Across all participants, each subset of
items was presented under both types of instruction, and the order in which each combination of
item subset and instruction type was presented
was counterbalanced.

LONG-TERM VISUAL REMEMBERING

Procedure
Participants were tested individually. It was
explained that as soon as they had in mind a
definite image of each named item they should
rate its vividness on the 5-point scale and then
draw the item in the rectangle below. They were
advised that when they had completed their
sketch they should turn the page and provide
ratings on a further two scales. On each trial
the time to evoke an image was logged by the
experimenter. With a simple key press the experimenter started an electronic counter as they
named the item to be imaged and drawn, and then
stopped the counter the moment they saw the
participant mark the image vividness scale.6

Results
Coding the drawings
Every participant produced a drawing for all
the named items. These were inspected by one of
the authors (L.M.) and for each item a comprehensive list was compiled of all the features
depicted in at least one drawing of the item. Two
independent judges (colleagues), who did not
know the purpose of the study, or of the conditions
under which any of the drawings had been
produced, worked through each printed list and
classified every feature as being either categorydefining or category-irrelevant. A categorydefining feature was described as being a feature
that is central to the items classification; that is,
something an item needs to possess, or typically
possesses, to qualify as belonging to the named
category.7 For a bag, for example, a pouch (with
opening) and a handle would be category-defining
features. Similarly, for a bed, a base unit, legs, and
headboard would be regarded as being categorydefining (although each individual leg would not
be counted separately). The agreement between

267

the judges with regard to which features were

category-defining was very high (Kappa.89).
On the few occasions where there was disagreement, a third independent judge was consulted.
The two judges then inspected the drawings and
identified which category-defining features were
depicted in each. Again there was a high level of
agreement with regard to which category-defining
features were present and, therefore, how many
such features were depicted in each drawing
(Kappa.91). A high level of agreement also
emerged when the two judges identified which
category-irrelevant features were depicted in
each drawing and, therefore, how many of these
features were in a drawing (Kappa .73). For
example, decoration of the outside of the pouch
of a bag would count as a single categoryirrelevant feature, as would decoration of the
handle. Similarly, a pillow and a blanket would
count as two category-irrelevant features of a
bed. Again, the independent judge resolved the
disagreements. Figure 2 illustrates the increased
tendency to depict category-irrelevant features
with personal exemplar instructions compared to
generic instructions. For the generic instructions
the illustrative drawings are seen to contain two
category-defining features in the case of the bag
(i.e., the pouch and handle), and three categorydefining features in the case of the bed (i.e., base,
legs, and headboard). For the personal exemplar
instructions the two illustrative drawings are seen
to contain additional category-irrelevant features,
generic

personal exemplar

Given the experimenter knew the condition under which

each named item was being imaged, it is acknowledged that
this is not an ideal procedure for protecting against experimenter bias. Fortunately, the effects on image evocation time
observed in the present study replicate the effects reported in
previous studies.
7
While having four legs or a headboard are not essential
features of a bed, beds normally have four legs and a
headboard. Although they can still be categorised as a bed
without these features, to do so requires special consideration,
and this might be reflected in a verbal qualification, such as
airbed.

Figure 2. Typical drawings generated in response to generic

(e.g., Please draw a bag) and personal exemplar (Please
draw your bag) instructions, illustrating the greater number
of category-irrelevant features depicted with the latter.

268

WALKER ET AL.

two in the case of the bag (i.e., decoration of pouch

and of handle) and three in the case of the bed
(i.e., cover, motif on cover, pillow case).
Analysis of variance
Participant-based analysis of variance (ANOVA) provided a preliminary assessment of the
impact that changing the instructions had on each
of the measures taken, including the number of
category-defining and category-irrelevant features. An alpha level of .05 was selected for all
statistical tests in the present study. Each ANOVA
had Instruction Type (generic versus personal
exemplar) and Serial Position (1
10) as withinparticipants factors, and Order (generic-personal
versus personal-generic) as a between-participants
factor. Serial Position refers to the position of an
item within a subset of to-be-drawn items. Because
items within a subset were presented in a random
order, that was different for each participant,
Order is not of central concern here. In no case
was there a main effect of Order, or an interaction
with Instruction Type, and no further reference is
made to these.
Image specificity. Mean ratings of image specificity were 1.70 (SEM0.11) and 4.14 (SEM 
0.09) for the generic and personal exemplar
instructions, respectively. The difference was
highly significant, F(1, 18) 312.4, p B.001, partial h2 .95. With lower values indicating less
specificity, the results confirm that the generic
instructions, despite allowing participants to generate images and drawings of specific items, did
induce participants to generate images that were
more generic than those induced by the personal
exemplar instructions.
Episodic reference. Participants images had
stronger episodic reference (i.e., referred to a
particular previous occasion on which they had
encountered the imaged item) following personal
exemplar instructions than following generic instructions (M2.52, SEM0.14, and M 1.67,
SEM0.07, respectively), F(1, 18)45.78, pB.001,
partial h2.72.8

At first blush a mean of 2.52 seems rather low. However,

the wording associated with ratings of 4 and 5 might have been
unrealistic for anything being retrieved from long-term
memory. These ratings didnt just require some aspect(s) of
a particular occasion to be imaged, they also required either
most aspects (score  4) or the whole (score  5) of a

Category-defining features. The number of

category-defining features depicted in a drawing
did not vary according to the type of instruction
(M 2.34, SEM0.04, and M 2.35, SEM
0.05, for generic and personal exemplar instructions, respectively), F(1, 18) B1.
Category-irrelevant features. There were fewer
category-irrelevant features depicted in drawings
produced under generic instructions than in drawings produced under personal exemplar instructions (M0.22, SEM0.03, and M0.93, SEM
0.08, respectively), F(1, 18)116.33, pB.001, partial h2 .87.
Time to evoke an image. The average time
to evoke an image was longer with personal exemplar instructions than with generic instructions (M6.22 s, SEM0.41, and M4.73 s,
SEM0.38, respectively), F(1, 18)14.45, p
.001, partial h2 .45.
Image vividness. Images generated in response
to personal exemplar instructions were rated
as being more vivid than images generated
in response to generic instructions (M 3.67,
SEM 0.06, and M 3.12, SEM0.08, respectively), F(1, 18) 32.32, p B.001, partial h2 .64.

Statistical modelling
These ANOVAs confirm the predictions regarding how personal exemplar instructions,
relative to generic instructions, would impact on
the measured variables. Unfortunately, however,
they do not reveal the pattern of interdependencies among the various measures taken. For
example, it remains unclear if the type of instruction influenced the depiction of category-irrelevant features directly, or only indirectly via its
effect on image specificity. Similarly, it remains
unclear if instruction type influenced the vividness of images directly, or only indirectly via its
effect on image specificity and/or the depiction of
category-irrelevant features. To gain a clearer
picture of the pattern of interdependencies

particular occasion to be imaged. In which case a mean score

of 2.52 begins to look different. It is worth noting, however,
that it is the relative scores observed for different conditions
that is most pertinent, and the statistical modelling to be
reported below treated the scores as such; that is, as ordered
categorical responses.

LONG-TERM VISUAL REMEMBERING

among the various measures taken, a series of

statistical models was fitted.
Preliminary explanatory variables. The continuous covariate Age, and the two binary variables
Gender (0 male, 1 female) and Order (0
personal-generic, 1generic-personal) were treated as between-participants factors. The binary
variable Instruction Type (0generic, 1
personal) was the only within-participants factor.
To test whether the effect of Instruction Type
varied with Order, the first-order interaction
variable Order-By-Instruction-Type was created.
Other first-order interactions could have
been examined but were not considered in this
analysis. The variables Age, Gender, Order, Instruction Type, and Order-By-Instruction-Type
constituted the set of preliminary explanatory
variables.
Ordinal categorical responses. The variable Image Specificity was classified as an ordinal categorical response repeated across 20 different images
for each participant. A repeated (clustered) ordered logit or cumulative logit or proportional
odds model (Agresti & Natarajan, 2001) was used
to relate Image Specificity to the set of preliminary
explanatory variables. The ologit command in
STATA was used with the cluster(ID) option
(Rabe-Hesketh & Skrondal, 2008; STATA, 2008)
to produce robust standard errors for the parameter estimates. The variables Episodic Reference
and Image Vividness were also regarded as repeated ordered categorical responses. A repeated
ordered logit model was used to examine the
effects of Image Specificity on Episodic Reference,
while controlling for the preliminary explanatory
variables. The same type of model also was used to
explore the impact of Image Specificity, Episodic
Reference, Number of Category-Defining Features, and Number of Category-Irrelevant Features on Image Vividness, each time adjusting for
the preliminary explanatory variables.
Binary responses. The number of categorydefining features and the number of categoryirrelevant features were both reclassified as binary
responses repeated over the 20 different images
for each participant. The number of categorydefining features was recoded according to
whether it was B3 (coded as 0) or 2 (coded as
1). The number of category-irrelevant features was
recoded according to whether it was B1 (coded as
0) or 0 (coded as 1). For each binary response
separately, a repeated logit model (or repeated

269

binary logistic regression) (Agresti, 2002) was

fitted using the ologit command in STATA with
the cluster(ID) option (Rabe-Hesketh & Skrondal,
2008). The effects of Image Specificity were
estimated in the presence of the preliminary
explanatory variables.
Durations. The time interval logged by the
experimenter was considered as a duration repeated across the 20 images. A Cox proportional hazards model with shared frailty (Kelly,
2004) was fitted using the stcox command in
STATA with the share(ID) option (Cleves, Gould,
Gutierrez, & Marchenko, 2008). This model
was used to explore the effects of Image Specificity, Episodic Reference, Number of CategoryDefining Features, and Number of CategoryIrrelevant Features on the Time to Evoke an
Image, each time controlling for the preliminary
explanatory variables.
Outcomes of modelling. Item specificity: The
model containing all the preliminary explanatory
variables was significant, Wald x2 (5, N400)
57.45, pB.001, and explained 19% (pseudo R2)
of the variation in Image Specificity. However,
the only significant predictor of Image Specificity
was Instruction Type (Odds Ratio33.6, z5.16,
pB.001), confirming that Image Specificity was
higher (i.e., imaged objects were more specific)
with personal exemplar instructions than with
generic instructions.
Episodic reference: The model incorporating
Image Specificity and the preliminary explanatory
variables was significant, Wald x2 (9, N400)
89.86, pB.001, and explained 12% (pseudo R2)
of the variation in Episodic Reference. The only
significant predictor of Episodic Reference was
Image Specificity. When the impact of successive
levels of Image Specificity were considered,
relative to the level of least specificity, the
Odds Ratios increased steadily (Odds Ratio1.8,
6.7, 7.1, 10.9, z1.86, 4.50, 4.15, 3.72, and
p.06,B.001,B.001,B.001, respectively), confirming that increasing levels of Image Specificity predicted higher levels of Episodic Reference.
To illustrate, the highest level of Image Specificity was associated with close to an eleven-fold
increase in the extent of Episodic Reference
compared to the extent associated with least
specificity.
Category-defining features: The model incorporating Image Specificity and the preliminary explanatory variables was marginally
significant, Wald x2 (9, N 400)16.13, p .06,

270

WALKER ET AL.

accounting for less than 1% of the variation in the

Number of Category-Defining Features. None of
the variables contributed significantly to the
model.
Category-irrelevant features: The model incorporating Image Specificity and the preliminary
explanatory variables was highly significant, Wald
x2 (9, N 400) 595.75, p B.001, explaining 17%
of the variation in the Number of CategoryIrrelevant Features. Only Instruction Type contributed significantly to the model (Odds
Ratio6.97, z8.01, pB.001), confirming that
drawings produced under personal exemplar instructions were almost seven times more likely to
depict at least one category-irrelevant feature
than were drawings produced under generic
instructions.
Time to evoke an image: Two variables from
the set of preliminary explanatory variables were
the only variables to impact significantly on
Image Evocation Time. Adding other variables
to the model failed to increase its explanatory
power even though, when considered alone, some
appeared to have a significant impact on Image
Evocation Time (e.g., extent of episodic reference
and number of category-irrelevant features). The
model incorporating just the preliminary explanatory variables was significant, Wald x2
(5, N 400) 56.78, p B.001. Instruction Type
(Hazard Ratio0.4; z 6.07; p B.001) and
Age (Hazard Ratio 0.97; z  2.54; p .011)
both impacted significantly on Image Evocation
Time, with images taking longer to evoke in
response to personal exemplar instructions than
in response to generic instructions, and as the age
of the participants increased.
Image vividness: Other than Instruction Type,
only the number of Category Irrelevant Features
impacted significantly on Image Vividness. The
final model that included this variable along with
the preliminary explanatory variables was significant, Wald x2(6, N 400) 57.09, pB.001, explaining 6% (pseudo R2) of the variation in Image
Vividness. More vivid images were induced
by personal exemplar instructions than by generic instructions (Odds Ratio 2.50, z3.10,
p.002). Images containing at least one category-irrelevant feature also were more vivid than
images with no such features (Odds Ratio1.46,
z1.99, p.046).
Figure 3 summarises the pattern of interdependencies revealed by the regression analyses.

instruction type

age

image evocation
time
specificity of
image

number of cat.
defining features
vividness

number of cat.
irrelevant features

episodic reference to
a previous occasion

Figure 3. The interdependencies among different measured

aspects of imaging and drawing revealed through regression
analyses of the consequences of adopting personal exemplar
instructions rather than generic instructions in Experiment 1.
Continuous lines depict influences attributed to imaged object
specificity, whereas dashed lines depict influences attributed to
other factors collectively identified as personal relevance.

Discussion
Personal exemplar instructions, compared to
generic instructions, made it more likely that
specific individual objects, rather than generic
objects, would be imaged. Mediated by this
increase in the specificity of the imaged objects,
personal exemplar instructions also introduced an
element of episodic reference, with participants
being more likely to confirm that each of their
images related to a particular occasion on which
the imaged object had been encountered. As
Figure 3 shows, instruction type also impacted
directly on other aspects of imaging and drawing.
Independent of their impact on image specificity,
personal exemplar instructions increased the
vividness of the images. They did this both
directly and indirectly by increasing the presence
of category-irrelevant features. It would appear,
therefore, that the presence of category-irrelevant
features, rather than of episodic reference, enhances image vividness. As expected, neither
object specificity nor personal relevance, as distinct aspects of the contrast in type of instruction,
influenced the number of category-defining features depicted in a drawing.
It would appear that personal relevance and
object specificity can each influence the type
of object information that is remembered. The
personal relevance of an object directly enhances
memory for its unique (category-irrelevant)
features. It can do so without reference to
any episodic information relating to previous

LONG-TERM VISUAL REMEMBERING

encounters with the object. On the other hand,

the more specific (less generic) the object information being retrieved, the more relevant are the
details (e.g., the spatio-temporal coordinates) of
previous encounters with this named object, and
the more likely are these details to be retrieved.
Conversely, of course, episodic information relating to encounters with specific objects is not
retrieved as part of the generic information
about objects (i.e., information pertinent to their
categories).
Finally, images can take longer to evoke in
response to personal exemplar instructions compared to generic instructions. This is consistent
with previous research showing that images of
specific items take longer to evoke than images of
generic items, and that images with episodic
reference take longer to evoke than images
lacking such reference (Cornoldi et al., 1989; De
Beni & Pazzaglia, 1995; Gardini et al., 2004;
Kosslyn, 1981; Palladino & De Beni, 2003).
Because in the present study the time to evoke
an image was directly dependent on instruction
type, without any mediation from object specificity, it would appear that personal relevance was
primarily responsible for the extra time needed
to generate an image with personal exemplar
instructions.

EXPERIMENT 2: ELABORATING THE

NATURE AND DETERMINANTS OF
EPISODIC REFERENCE
In Experiment 1 the effect of instruction type on
episodic reference was mediated entirely by the
specificity of the imaged object: The more generic
the imaged object, the less reference there was to
episodic information. The absence of any link
between personal relevance and episodic reference is a little surprising given the importance of
keeping track of where personally relevant items
are to be found and what state they are in.
Episodic memories preserve information about
when and where individual items have previously
been encountered, with where including reference
to other objects in the scene at the time. In
Experiment 1 the involvement of episodic memory was indexed by a single question concerning
the extent to which an image of a named item
related to a particular occasion on which the item
had been encountered. Unfortunately the term
occasion leaves some uncertainty regarding which

271

aspect of episodic reference is most salient (i.e.,

the extent to which an occasion is defined by the
when and/or the where).
Experiment 2 was designed to provide additional information regarding the separate influence of object specificity and personal relevance
on the when and where of episodic visual memory.
It was predicted that some aspects of episodic
reference would be linked directly to each of
these factors. Specifically, given the importance of
knowing where personally relevant objects can be
found, it was expected that personal relevance
would be linked to the retrieval of information
about the location of the previous encounter on
which the image of an object is based.

Method
The method was essentially the same as that
adopted in Experiment 1. However, the focus of
interest was limited to instruction type and image
specificity, which were the only significant predictors of episodic reference in Experiment 1.

Participants
The sample comprised 20 students and administrative staff at Lancaster University (5 males, 15
females) with a mean age of 21.8 years (SD 8.0
years). No payment was given for their participation, although the students received course credit.

Materials
There was just one change to the list of 20 tobe-drawn items. This involved replacing belt, for
which relatively few category-irrelevant features
were offered, with watch.
Image specificity. On the reverse side of each
response sheet was a slightly modified 5-point
scale designed to assess the extent to which the
completed drawing was based on an image that
was typical for the objects category, rather than
on an image of a particular individual object
(1completely typical, not at all a particular
individual object, 2mostly typical, slightly a
particular individual object, 3an equal mix of
a typical and of a particular individual object,
4slightly typical, mostly a particular individual
object, 5 not at all typical, completely a particular individual object).

272

WALKER ET AL.

In addition there were three new questions to

be answered concerning the when and where of
any previous encounter that provided the memory representation on which the image was based.
Episodic reference to a specific time and
recency. The first of these new questions required
participants to indicate the extent to which their
image of the named object was linked to a specific time when the object was encountered
(1clearly from a specific time, 2somewhat
from a specific time, 3not at all from a specific
time). The second new question required participants to provide an estimate of the recency of
their encounter with the object. Specifically, they
were asked to indicate how many times they had
seen the named object since the specific time
(encounter) from which their image came. Participants were asked to provide such an estimate in
those cases where they had indicated their image
of the named item was either clearly from a
specific time or somewhat from a specific time.
Episodic reference to a specific place. The third
new question required participants to indicate
the extent to which their image of the named
object was linked to a specific place in which
the object was encountered (1 clearly from a
specific place, 2somewhat from a specific place,
3not at all from a specific place).

Apart from these changes, the method was the

same as for Experiment 1.

Results
Coding the drawings
Every participant produced a drawing of all the
named items. These were inspected by one of the
authors (H.K.) and an independent judge (i.e., a
colleague who did not know the purpose of the
study, or of the conditions under which any of the
drawings had been produced), and for each item a
note was made of any objects, other than the
named object, that were depicted in the drawing.
Subsequently, where there was lack of agreement,
this was resolved through discussion. Figure 4
shows two typical drawings produced under
personal exemplar instructions in which additional objects were included.
Analysis of variance
Participant-based ANOVA provided a preliminary exploration of the impact of instruction

Episodic reference: Inclusion of additional

objects in the image. The where context for an
image was also assessed with a question that
asked participants to name any other objects in
their image. Any such other objects were to be
regarded as providing the scenic context for
the encounter (i.e., the objects were alongside
the named object in the particular encounter on
which their image was based).
Episodic reference: Inclusion of additional
objects in the drawing. Finally, at the conclusion
of the experiment the experimenter and an independent judge identified any further additional
objects that were depicted in the drawing of each
named object.9
9
Although the theoretical significance of the presence of
other objects in participants images and drawings was not
appreciated in advance of Experiment 1, so their presence was
neither indexed nor analysed, it was noticed that some of the
drawings incorporated such objects. Rather than conduct a
post hoc analysis to examine their presence as an independent
variable, it was decided to derive theoretically motivated
expectations relating to their presence in preparation for
Experiment 2.

Figure 4. Typical drawings generated in response to personal

exemplar instructions in Experiment 2, illustrating the inclusion of other objects from the encounter on which the drawing
of the named object is based. The drawings were produced in
response to the instructions Please draw your TV and
Please draw your bed.

LONG-TERM VISUAL REMEMBERING

type on each of the measures taken. Each

ANOVA had Instruction Type and Serial Position as within-participants factors. No betweenparticipants factors were included.
Image specificity. Mean ratings of image
specificity were 2.24 (SEM 0.14) and 3.89
(SEM0.19) for the generic and personal exemplar instructions, respectively. The difference
was highly significant, F(1, 19) 101.3, pB.001,
partial h2 .84. With lower values indicating less
specificity, the results confirm that the generic
instructions induced participants to generate
images that were more generic than those induced
by the personal exemplar instructions.
Episodic reference to a specific time. Participants images were associated with a stronger
sense of being linked to a specific time following
personal exemplar instructions than following
generic instructions (M 2.02, SEM 0.13, and
M 2.53, SEM 0.07, respectively), F(1, 19) 
14.54, p .001, partial h2 .43. Figure 5 also
confirms the presence of a recency effect linked
to participants claims that their images were
clearly from a specific time. When participants
made this claim their image was most likely to be
based on their most recent encounter with the
named item, a trend that was less evident when
they claimed their image was only somewhat from
a specific time.
60
clearly from a specific time
somewhat from a specific time

50

frequency

40
30
20
10
0
0

>=4

number of encounters with the imaged object

since the encounter on which the image is based

Figure 5. The recency effect that emerged when participants

in Experiment 2 confirmed that their drawings were based on
encounters with a strong sense of being linked to a specific
time. Note the more pronounced recency effect when the
image was clearly from a specific time as opposed to
somewhat from a specific time.

273

Episodic reference to a specific place. Participants images were associated with a stronger
sense of being linked to a specific place following
personal exemplar instructions than following
generic instructions (M 1.43, SEM0.07, and
M 2.30, SEM 0.07, respectively), F(1, 19) 
97.76, p B.001, partial h2 .84.
Episodic reference: Inclusion of additional
objects in the image. Participants images were
more likely to incorporate an additional object
following personal exemplar instructions than following generic instructions (M1.31, SEM
0.15, and M0.50, SEM0.08, respectively),
F(1, 19)55.43, pB.001, partial h2 .75.
Episodic reference: Inclusion of additional
objects in the drawing. In general, fewer additional objects were included in the drawings
than in the images, suggesting either some
editing or some loss of information as the image
was translated into a drawing. Nevertheless
participants drawings, like their images, were
more likely to incorporate an additional object
following personal exemplar instructions than
following generic instructions (M 0.93, SEM 
0.10, and M 0.28, SEM0.05, respectively),
F(1, 19) 61.05, pB.001, partial h2 .76.
Statistical modelling
Preliminary explanatory variables. The continuous covariate Age, and the two binary variables Gender (0male, 1 female) and Order
(0personal-generic, 1generic-personal) were
again treated as between-participants factors. The
binary variable Instruction Type (0 generic,
1personal) was the only within-participants
factor. First-order interactions were not considered in this analysis. The variables Age, Gender,
Order, and Instruction Type constituted the set of
preliminary explanatory variables.
Ordinal categorical responses. The variable
Image Specificity was again classified as an
ordinal categorical response repeated across 20
different images for each participant (cf. Experiment 1). A repeated (clustered) ordered logit or
cumulative logit or proportional odds model was
used to relate Image Specificity to the set of
preliminary explanatory variables. The variables
Specific Time and Specific Place were also
regarded as repeated ordered categorical responses. A repeated (clustered) ordered logit
model was used to examine the effects of Image

274

WALKER ET AL.

Specificity on these two variables, while controlling for the preliminary explanatory variables.
Count data. The number of other objects
included in an image of a named item (i.e.,
Additional Objects) was treated as a count
variable and referred to the Poisson distribution
for analysis. The poisson command in STATA was
used to relate this variable to the preliminary
variables and Image Specificity, with the cluster(ID) option again being used to produce robust
standard errors for the parameter estimates.
Outcomes of modelling. Image specificity:
The model containing all the preliminary explanatory variables was significant, Wald x2(5,
N 400) 105.94, pB.001, and explained 11%
(pseudo R2) of the variation in Image Specificity.
However, the only significant predictor of
Image Specificity was Instruction Type (Odds
Ratio9.33, z8.00, pB.001), confirming that
Image Specificity was higher (i.e., imaged objects
were more specific) with personal exemplar
instructions than with generic instructions.
Episodic reference to a specific time: The
model incorporating Image Specificity and the
preliminary explanatory variables was significant,
Wald x2(9, N 400) 39.74, p B.001, and explained 14% (pseudo R2) of the variation in
Specific Time. The only significant predictor of
Specific Time was Image Specificity. When the
impact of successive levels of Image Specificity
were considered then, relative to the level of least
specificity, the Odds Ratios decreased steadily
(Odds Ratio0.22, 0.09, 0.07, 0.08, z  3.37,
4.59, 3.92, 3.90, all pB.001), confirming
that increasing levels of Image Specificity predicted an increasing sense that an image was
linked to a specific time. To illustrate, the highest
level of Image Specificity was associated with
close to a 13-fold increase in the strength of the
sense that a specific time was present, compared
to the strength associated with least specificity.
Episodic reference to a specific place: The
model incorporating Image Specificity and the
preliminary explanatory variables was significant,
Wald x2(9, N400) 131.79, p B.001, and explained 23% (pseudo R2) of the variation in
Specific Place. Instruction Type and Image Specificity were both significant predictors of Specific
Place. The Odds Ratio for Instruction Type was
0.27 (z 4.23, p B.001), confirming that personal exemplar instructions, relative to generic
instructions, induced a greater sense that the
image was linked to a specific place. When the

impact of successive levels of Image Specificity

were considered, relative to the level of least
specificity, the Odds Ratios decreased steadily
(Odds Ratio 0.20, 0.11, 0.05, 0.05, z 3.94,
4.78, 4.44, 4.85, all p B.001), confirming
that increasing levels of Image Specificity predicted an increased sense that an image was
linked to a specific place. The highest level of
Image Specificity was associated with a 20-fold
increase in the strength of the sense that a specific
place was present, compared to the extent associated with least specificity.
Episodic reference: Inclusion of additional
objects in the image. The model incorporating
Image Specificity and the preliminary explanatory variables was significant, Wald x2(9,
N 400) 353.91, p B.001. Instruction Type was
the only significant predictor of Additional Objects, Incidence-Rate Ratio (IRR) 2.37, z 5.82,
p B.001, confirming that it had a direct influence
on the retrieval of additional objects, irrespective
of how generic or specific was the information on
which the image was based. Successive levels of
Image Specificity were not associated with an
increasingly significant impact on the number of
additional objects. For only one of the four
successive levels of change in Image Specificity
was there a significant impact on the number of
additional objects (IRR1.44, 1.63, 1.51, 1.70,
z 1.95, 2.05, 1.46, 1.74, p .052, .04, .15, .08).
Figure 6 summarises the pattern of interdependencies revealed by the regression analyses.

Discussion
Experiment 2 reveals the different ways in which
personal relevance and object specificity enhance
episodic reference. All three aspects of episodic
reference were enhanced when personal exemplar instructions, rather than generic instructions,
were adopted, and statistical modelling confirmed
the separate contributions from object specificity
and personal relevance.
With regard to episodic reference to a specific
time, the impact of instruction type was mediated
entirely through its influence on image specificity.
Images of specific individual objects, compared to
generic images, were accompanied by a stronger
sense of a specific time at which an imaged object
had previously been encountered. Furthermore,
as expected on the grounds that the most recent
of previous encounters is especially salient

LONG-TERM VISUAL REMEMBERING

275

same way as reference to an occasion in Experiment 1 (i.e., both were driven entirely by image
specificity, and not at all by instruction type
directly), it would appear that to make reference
to an occasion can be, in effect, to make reference
just to the sense of time linked to an encounter.

instruction type

specificity of
image

image referred to
specific time (with a
strong element of
recency)

number of
other objects
in image
image
referred to
specific place

Figure 6. The interdependencies among different aspects of

imaging revealed through regression analyses of the consequences of adopting personal exemplar instructions rather
than generic instructions in Experiment 2. Continuous lines
depict influences attributed to imaged object specificity,
whereas dashed lines depict influences attributed to personal
relevance.

because it is most likely to be an objects current

location, there was a strong recency effect associated with images of specific objects: It was the
most recent encounter with a specific object that
tended to provide the basis for an image of the
object (see Figure 5).
Instruction type had both a direct effect on the
sense that an image was linked to a specific place,
and an indirect effect via image specificity. It
appears, therefore, that both personal relevance
and object specificity increase the sense that the
information being retrieved from long-term memory relates to a particular spatial location in which
the object had been encountered.
Finally, adopting personal exemplar instructions, rather than generic instructions, induced the
inclusion of other objects in the images and
drawings of the named items. Because this was
independent of object specificity, it is attributed
to personal relevance. Looking across both experiments, the functional significance of personal
relevance appears to be the same with regard to
the inclusion of other objects (i.e., objects providing the scene for a previous encounter) and with
regard to the inclusion of an objects categoryirrelevant features (see Experiment 1). In both
cases the enhancements occur without reference
to the spatio-temporal aspects of previous encounters (i.e., an episodic sense of time and
place).
Finally, given that it was only episodic reference to a specific time that behaved in the

GENERAL DISCUSSION
There are many ways in which an object can
become personally relevant to an individual, and
presuming ownership of the object is just one of
these ways (Cunningham et al., 2008; Van den Bos
et al., 2010). However it is achieved, enhancing
the personal relevance of an item brings with
it several other changes, any of which could
contribute to the impact personal relevance has
on object memory. For example, it is normally
specific individual objects that become personally
relevant (rather than categories of object) (see
Van den Bos et al., 2010), and this enhanced
object specificity is very likely to impact on
memory. Employing a task situation in which
participants were asked to retrieve visual information about familiar objects from long-term
memory, the present study sought to confirm the
separate contributions to object memory of object
specificity and personal relevance.
It is argued that object specificity and personal
relevance should influence what is remembered
about our experiences of objects, and not just how
well the same information is remembered. For
example, the category-irrelevant features identifying specific, personally relevant objects need to
be remembered, along with a range of other facts,
such as what state the objects are in and where
they can be found. Details regarding the latter
should include the relative positions and identities
of other objects involved in previous encounters,
along with information about the time of each
previous encounter with a particular object, at
least sufficient to allow its most recent location to
be distinguished from earlier locations (which is
most likely to be its current location). On this
basis memory for specific, personally relevant
objects, and for more specific objects regardless
of personal relevance, should include more elements of the spatio-temporal context in which the
objects were encountered. The presumed increasing involvement of visual episodic memory in
remembering specific, personally relevant objects
is entirely consistent with Van den Bos et al.s
(2010) observation that enhancing personal

276

WALKER ET AL.

relevance, through ownership, increases the likelihood that participants will have a conscious
recollection of having previously seen an object
(i.e., they will remember the object rather than
just know they have seen it before). More
generally, this observation provides strong indication that personal relevance, and the object
specificity it entails, changes what we remember
about objects, rather than just how well we
remember them.
To identify the separate and distinctive contributions of object specificity and personal relevance to object memory, the two experiments
reported here employed a task in which participants imaged and drew familiar named objects
from long-term memory. They did this in response
to two different types of instruction, one type
encouraging them to image and draw generic
objects (e.g., Please draw a bed), and another
type encouraging them to image and draw specific
objects of personal relevance to them (e.g.,
Please draw your bed). By asking participants
to indicate how specific versus generic the object
in each of their images was, statistical modelling
(based on multiple regression) successfully dissociated the influence each of the two factors had
on various aspects of imaging and drawing from
long-term visual memory.
By contrasting the images and drawings generated in response to generic and personal
exemplar instructions, it was confirmed that the
object specificity that is induced by enhanced
personal relevance has its own impact on object
memory, separately from the impact that personal relevance otherwise has. These two factors
impacted in different ways on the types of
information about named objects that are retrieved from long-term visual memory. Personal
relevance induced the retrieval of a named
objects category-irrelevant features (i.e., the
unique features identifying that specific object)
and the retrieval of the other objects with which
it had previously been encountered (i.e., objects
providing the context for a particular encounter).
By increasing the presence of category-irrelevant
features of the object, personal relevance also
enhanced the vividness of the image of an object.
On the other hand, object specificity induced a
greater sense that the information being retrieved
is linked to a specific time in the past, with the
most recent encounter with an object being the
most salient. Personal relevance and object
specificity both enhanced the sense that the
information being retrieved is linked to a parti-

cular place. The corollary of all of this was that

retrieving generic information tended not to
include information about the timing of any
relevant previous encounters or of episodic
features more generally.
These are novel findings, which extend previous work by revealing the contributions to
object memory of one distinct factor thought to
co-vary with other factors as an aspect of personal
relevance. The present study also demonstrates
the appropriateness and power of one approach
to statistical modelling that could be usefully
applied to isolate other factors contributing to
personal relevance. Furthermore, by shifting the
focus away from examining how temporary ownership in an experimental situation can impact on
temporary (i.e., relatively short-term) memory
(see Cunningham et al., 2008; Van den Bos
et al., 2010), and towards the long-term consequences of personal relevance for remembering
objects encoded in everyday life, the present
study extends the generality and ecological validity of previous work. Finally the present study
incorporates some potentially fruitful links with
other work, from which a broader theoretical
perspective can be brought to bear.
It is proposed here that generic visual information lacking personal relevance is retrieved from
visual knowledge, whereas visual information
about specific objects having personal relevance
is retrieved from visual episodic memory (what
Kosslyn, 1994, refers to as the category pattern
activation and exemplar pattern activation systems
respectively).10 Visual knowledge is generally
regarded as incorporating information specifying
the prototypical visual features of item categories
at varying levels of abstraction (i.e., the visual
knowledge base explored by Rosch et al., 1976).
It is to this type of visual representation that
other category information is linked, including
the name of the category. Visual episodic memory
is thought to preserve records of encounters with
particular objects, allowing the encounters to be
re-experienced. Thus episodic memories preserve
information identifying the particular object
encountered (i.e., including its category-irrelevant
features) and the context in which it was
10

There is neuropsychological evidence that visual knowledge and visual episodic memory are distinct cognitive
resources. For example, depending which regions of the brain
are damaged through injury, either one of these sources
of information can be selectively degraded (Kapur, 1999;
Tulving, 2002; Wheeler & McMillan, 2001).

LONG-TERM VISUAL REMEMBERING

encountered; that is, the when and where of the

encounter, with the latter including the other
objects providing a context for the encounter.11
The contributions of visual knowledge and
visual episodic memory are not thought to be
mutually exclusive, but instead are thought to be
capable of contributing simultaneously to the
same visual image. This is implied by the view
that visual images can have a mix of objectspecific and generic elements, which in turn is
confirmed by previous observations that images
of named items are initially generic in nature and
only then are elaborated to become more specific
(Cornoldi et al., 1989; De Beni & Pazzaglia, 1995;
Kosslyn, 1981) and more episodic (Gardini et al.,
2004; Palladino & De Beni, 2003). It would seem,
therefore, that at some intermediate point in their
generation images will reflect the combined
influence of visual knowledge and visual episodic
memory. This account also explains why specific
images were observed to take longer to evoke
than generic images in Experiment 1.
Episodic memory seems to be especially vulnerable to the effects of increasing age (see Light,
1991, for a review). If this means episodic
memory becomes increasingly difficult to use,
then we have an explanation for the effect of
age on image evocation times in Experiment 1.
When explicitly instructed to generate an image
of a specific object that is personally relevant, any
participants finding this difficult to do will need
extra time to do so. Although Palladino and De
Beni (2003) observed specific images to be less
frequently generated with age, relative to generic
images, the present findings suggest it might be
important to examine further whether it is specificity that is the key feature. The effects of age on
image evocation time were independent of object
specificity and could be reflecting difficulties in
accessing visual episodic memory generally.
People suffering either from major depressive
disorder (see Williams & Scott, 1988, and Williams et al., 2007, for a review), or from posttraumatic stress disorder (Schonfeld & Ehlers,
2006), are two additional groups who find it
difficult to access memories of specific episodes
in their life that occurred at a particular time and
place on a particular day. Instead they retrieve
generic information about a category of event.
For example, in response to the cue word party, or
11
The context in which an object was encountered will also
include how it was encountered (e.g., its orientation from the
viewpoint from which it was seen).

277

an equivalent picture, they might respond by

saying I always enjoy a good party but not by
saying I enjoyed Janes party last Friday (cf.
Williams et al., 2007). In light of the evidence
from the present study, and the theoretical framework within which it is placed, it is predicted that
the visual images and drawings of these same
client groups will show much less sensitivity to the
requirements associated with personal exemplar
instructions compared to generic instructions.
That is, they will tend to continue to rely on
visual knowledge and the generic information this
makes available. If this is confirmed, then the
features of images and drawings that signal the
differential involvement of visual knowledge and
visual episodic memory could provide useful
confirmation as to when emotional disturbance
is impacting on a persons ability to access
information preserved in their episodic memory.
This would be particularly useful, of course, for
client groups who are unable or unwilling to
communicate verbally with the levels of fluency
required when depression is being assessed in
more conventional ways.
Finally, the present results resonate with those
from other studies in which drawing has been used
to identify the circumstances under which generic
and object-specific episodic information are preferentially retrieved from visual memory (Walker,
Blake, & Bremner, 2008; Walker, Bremner, Smart,
Pitt, & Apsey, 2008; Walker, Kennedy, & Berridge,
2011). In these studies, in which participants had
to draw a small number of objects from recent memory, Walker and his colleagues gave
themselves the opportunity to manipulate the
availability of generic visual information (i.e.,
information about object categories) by using
objects that were novel. The creation and utilisation of generic information was then controlled by
arranging for each object either to be named at
encoding with a novel count noun (This is a dax)
or to not be so named (This is an object), the
assumption being that only in the former case
would generic information about the objects be
established (see Walker, Bremner, et al., 2008, for a
review of evidence justifying this assumption).
When the objects were named at encoding, and
no constraints were imposed on how they should be
drawn, participants revealed a bias towards drawing them in a way that reflected their generic
aspects, rather than in a way that reflected
how they had been experienced. Specifically, the
objects were drawn in unseen prototypical orientations, rather than in the orientations in which they

278

WALKER ET AL.

had been seen. However, when the objects were

not named at presentation, so that generic representations were not available, they tended to be
drawn in the orientations in which they had been
seen.12
Walker and his colleagues propose that
although information about how individual objects had previously been experienced is preserved in long-term visual memory, people
prefer to make use of generic information about
the objects. In other words, the bias to retrieve
generic information about objects from long-term
visual memory is a matter of choice that reflects
the relative accessibility, rather than availability,
of generic and episodic representations. Evidence
supporting this proposal was obtained from a
condition in which participants were asked to
draw named objects specifically in the orientations in which they had been seen. Despite the
bias towards prototypical orientations that would
normally be in evidence, participants were able to
draw the objects in the orientations in which they
had seen them (Walker et al., 2011).
To summarise, therefore, in the context of
much longer-term memory than was examined
by Walker and his colleagues, the results of the
present study support their proposal that generic
and object-specific episodic memory representations can co-exist in long-term memory, and that a
difference in their accessibility, rather than their
availability, normally induces a bias towards
relying on the former whenever this is acceptable.
The results of the present study confirm that
other types of information linked to visual
episodic memory also will be retrieved when
circumstances discourage reliance on generic
information. Thus, in addition to the orientation
in which an object was previously encountered,
discouraging reliance on generic representations
will induce the retrieval of information providing
a sense of time and place, information regarding
an objects unique features, and information
about other objects encountered at the same time.
Manuscript received 5 December 2011
Manuscript accepted 28 August 2012
First published online 21 September 2012

12

The one exception to this general rule related to the final

novel object encountered during encoding which, if drawn
first, was drawn in the orientation in which it was seen,
regardless of whether it had been named.

REFERENCES
Agresti, A. (2002). Categorical Data Analysis (2nd ed.).
Hoboken, NJ: Wiley.
Agresti, A., & Natarajan, R. (2001). Modeling
clustered ordered categorical data: A survey. International Statistical Review, 69, 345
371.
Beggan, J. K. (1992). On the social nature of non-social
perception: The mere ownership effect. Journal of
Personality and Social Psychology, 62, 229
237.
Belk, R. W. (1988). Possessions and the extended self.
Journal of Consumer Research, 15, 139
168.
Belk, R. W. (1991). The ineluctable mysteries of
possessions. Journal of Social Behavior and Personality, 6, 17
55.
Bozeat, S., Lambon Ralph, M. A., Graham, K. S.,
Patterson, K., Wilkin, H., Rowland, J., . . .
Hodges, J. R. (2003). A duck with four legs:
Investigating the structure of conceptual knowledge
using picture drawing in semantic dementia. Cognitive Neuropsychology, 20, 27
47.
Cleves, M. A., Gould, W. W., Gutierrez, R. G., &
Marchenko, Y. U. (2008). An introduction to survival
analysis using Stata (2nd ed.). College Station, TX:
Stata Press.
Conway, M. A., & Dewhurst, S. A. (1995). The self and
recollective experience. Applied Cognitive Psychology, 9, 1
19.
Cornoldi, C., De Beni, R., & Pra Baldi, A. (1989).
Generation and retrieval of general, specific and
autobiographic images representing concrete nouns.
Acta Psychologica, 72, 25
329.
Cunningham, S. J., Turk, D. J., Macdonald, L. M., &
Macrae, C. N. (2008). Yours or mine? Ownership
and memory. Consciousness and Cognition, 17,
312
318.
De Beni, R., & Pazzaglia, F. (1995). Memory for
different kinds of mental images: Role of contextual
and autobiographic variables. Neuropsychologia, 33,
1359
1371.
Gardini, S., De Beni, R., & Cornoldi, C. (2004). Can we
have an image of a concept? The generation process
of general and specific mental images. Imagination,
Cognition and Personality, 23, 193
200.
Hoffmann, J., Denis, M., & Ziessler, M. (1983).
Figurative features and the construction of visual
images. Psychological Research, 45, 39
54.
Kahneman, D., Knetsch, J. L., & Thaler, R. (1991). The
endowment effect, loss aversion and status quo bias.
Journal of Economic Perspectives, 5, 193
206.
Kapur, N. (1999). Syndromes of retrograde amnesia: A
conceptual and empirical synthesis. Psychological
Bulletin, 125, 800
825.
Kelly, P. J. (2004). A review of software packages for
analyzing correlated survival data. The American
Statistician, 58, 337
342.
Knetsch, J. L., & Sinden, J. A. (1984). Willingness to
pay and compensation demanded: Experimental
evidence of an unexpected disparity in measures of
value. Quarterly Journal of Economics, 99, 507
521.
Kosslyn, S. M. (1981). The medium and the message in
mental imagery: A theory. Psychological Review, 88,
46
66.

LONG-TERM VISUAL REMEMBERING

Kosslyn, S. M. (1994). Image and brain: The resolution

of the imagery debate. Cambridge MA: MIT Press.
Light, L. L. (1991). Memory and aging: Four hypotheses in search of data. Annual Review of Psychology,
42, 333
376.
Palladino, P., & De Beni, R. (2003). When mental
images are very detailed: Image generation and
memory performance as a function of age. Acta
Psychologica, 113, 297
314.
Rabe-Hesketh, S., & Skrondal, A. (2008). Multilevel
and longitudinal modeling using STATA. College
Station, TX: Stata Press.
Rogers, T. B., Kuiper, N. A., & Kirker, W. S. (1977).
Self-reference and the encoding of personal information. Journal of Personality and Social Psychology, 35, 677
688.
Rosch, E. R., Mervis, C. B., Gray, W. D., Johnson,
D. M., & Boyes-Braem, P. (1976). Basic objects in
natural categories. Cognitive Psychology, 8, 382
439.
Schonfeld, S., & Ehlers, A. (2006). Overgeneral memory extends to pictorial retrieval cues and correlates
with cognitive features in posttraumatic stress disorder. Emotion, 6, 611
621.
STATA. (2008). Data analysis and statistical software
release 10.0. Stata Corp, Stata Press, College Station,
Texas.
Symons, C. S., & Johnson, B. T. (1997). The selfreference effect in memory: A meta-analysis.
Psychological Bulletin, 121, 371
394.
Tulving, E. (2002). Episodic memory: From mind to
brain. Annual Review of Psychology, 53, 1
25.
Van den Bos, M., Cunningham, S. J., Conway, M. A., &
Turk, D. J. (2010). Mine to remember: The impact
of ownership on recollective experience. Quarterly
Journal of Experimental Psychology, 63, 1065
1071.

279

Walker, P., Blake, H., & Bremner, J. G. (2008). Object

naming induces viewpoint-independence in longerterm visual remembering: Evidence from a simple
object drawing task. European Journal of Cognitive
Psychology, 20, 632
648.
Walker, P., Bremner, J. G., Merrick, K., Coates, S.,
Cooper, E., Lawley, R., . . . Simm, R. (2006). Visual
mental representations supporting object drawing:
How naming a novel object with a novel count noun
impacts on young childrens object drawing. Visual
Cognition, 13, 733
788.
Walker, P., Bremner, J. G., Smart, L., Pitt, T., & Apsey,
D. (2008). Object categorisation, object naming, and
viewpoint-independence in visual remembering:
Evidence from young childrens drawings of a novel
object. Memory, 16, 626
636.
Walker, P., Kennedy, H., & Berridge, D. (2011). Object
naming induces reliance on orientation-independent
representations during longer-term, but not shortterm, visual remembering. Memory, 19, 809
824.
doi:10.1080/09658211.2011.613838
Wheeler, M. A., & McMillan, C. T. (2001). Focal
retrograde amnesia and the episodic-semantic
distinction. Cognitive, Affective, and Behavioral
Neuroscience, 1, 22
36.
Williams, J. M. G., & Scott, J. (1988). Autobiographical
memory in depression. Psychological Medicine, 18,
689
695.
Williams, J. M. G., Barnhofer, T., Crane, C., Hermans,
D., Raes, F., Watkins, E., et al. (2007). Autobiographical memory specificity and emotional disorder.
Psychological Bulletin, 133, 122
148.

Copyright of Memory is the property of Psychology Press (UK) and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission. However, users
may print, download, or email articles for individual use.