COGNITIVE
PSYCHOLOGY
15,
238-300 (1983)
Conscious
and Unconscious
Perception:
An Approach
to the Relations between Phenomenal
Experience
and
Perceptual Processes
ANTHONY
MRC Applied Psychology
J. MARCEL
Unit, Cambridge,
England
An approach to the relationships between conscious perception and nonconscious perceptual processes is outlined. Its basis is the rejection of the assumption
that phenomenal experience is identical to or is a direct reflection of representations yielded by perceptual processes. Nonconscious perceptual processes
automatically redescribe sensory data into every representational
form and to the
highest levels of description available to the organism. Such processes (a) provide
records of each resultant representation,
(b) produce perceptual hypotheses in
different domains, (c) activate related structures, and (d) affect analog aspects of
actions. Conscious perception requires a constructive
act whereby perceptual
hypotheses are matched against information recovered from records, and serves
to structure and synthesize that information recovered from different domains.
These processes are related to three aspects of phenomenal experience: awareness, unity of percepts, and selectivity.
Consciousness is seen as an attempt to
make sense of as much data as possible at the most functionally
useful level.
Explication
of the approach consists of (a) discussion of differences between
conscious and nonconscious representations and processes; (b) exposition of the
characteristics of the process of recovery; (c) a theory of central visual masking as
a consequence of temporal and spatial parsing involved in recovery, wherein
masking is seen as an aspect of the structural nature of consciousness whose goal
is event perception, and does not affect nonconscious perceptual processing; (d)
an interpretation
of various clinical neuropsychological
and normal phenomena in
terms of limitations and impairments in the processes of recovery and synthesis; (e)
reinterpretation
of several perceptual phenomena in terms of the recovery of information and of how nonconscious processes precede and affect consciousness.
INTRODUCTION
Much psychological investigation and theorizing rests on the Identity
Assumption, whereby the representations which constitute conscious experience are assumed to be the very same ones that are derived and used
in sensory and cognitive processing. Serious doubt was cast upon this
paradigm assumption in the experimental
paper which precedes this
(Marcel, 1983). The data therein suggest a functional separation of (a)
representations which result from analysis or processing of sensory data
or aspects of it and which
can influence
behavior,
from (b)
phenomenological
representations
which can be reflected upon or reI am grateful to Leslie Henderson for helpful discussion on a previous draft of this paper,
to Stephen Palmer and Earl Hunt for painstaking suggestions as to clarification,
and to
George Mandler, Alan Baddeley, and Mike Posner for encouragement.
238
OOlO-0285183 $7.50
Copyright
@ 1983 by Academic
Press, Inc.
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
239
ported or which serve as the basis for intentional actions. In this paper an
attempt will be made to sketch an approach to the relationship of conscious percepts to nonconscious perceptual processing. These proposals
do not depend on the experiments in the preceding paper, but they were
provoked partially by them and are encouraged by them.
Before going any further it may be useful to recapitulate the results of
those experiments. Experiment 1 showed that when target-pattern mask
SOAs (stimulus-onset asynchronies) are too brief to enable detection,
graphic and semantic similarity judgments can be made to the target. As
SOAs were reduced, detection, graphic, and semantic judgments were
precluded in that order; that is, in the opposite order to which perceptual
analysis is supposed to proceed. The recalcitrance of certain subjects
itself emphasizes the inadequacy of a technique designed to assess processing which requires people to address their knowledge directly. In
Experiment 2, when choice stimuli covaried negatively on graphic and
semantic similarity, subjects could no longer reliably make such judgments. This suggests that graphic and semantic similarity judgments were
made on the basis of passive orientation responses due to residual activation from the nondetected word, rather than on the basis of subjects’
ability to selectively address characteristics of the masked word. Experiments 3, 4, and 5 showed that when an indirect measure of perceptual
processing is used (associative effects of the undetected word on a subsequent task) all subjects show the effects of undetected stimuli. Experiment 4 showed that while the effect holds with central pattern masking, it
is not obtained with peripheral energy masking. Experiment 5 showed
that repetition of a word followed by a pattern mask increases the effect of
semantic association on a related word, but has no effect on awareness or
on the semantic content of forced guesses.
These experiments raise several issues, the most obvious being the
nature of masking, the relation of apparently automatic nonconscious
processing to conscious representations and intentional processes, and a
reevaluation of studies which have taken consciously based responses to
reflect perceptual processes in general. What follows will attempt to deal
speculatively with these and related issues. What will be provided is not a
model, but rather a way of looking at the processes involved in conscious
experience and in tasks. Very few of the constructs in the theory are
novel; it is in their functional role that there is a change in emphasis.
To begin with an attempt will be made to indicate and give some criteria
for what is under discussion. A brief outline will then be given of the
proposed approach to the relations between conscious experience and
perceptual processing, primarily in vision but extendable to other domains. The main body of the paper consists of an expanded explication of
this approach. First, differences between nonconscious and conscious
240
ANTHONY
J. MARCEL
processes and representations will be discussed. Second, characteristics
of a central construct relating the two domains, the process of recovery of
information, will be dealt with. Third, an approach to central masking will
be put forward, whereby it is seen not to affect perceptual processing
itself, but as a phenomenological consequence of temporal and spatial
parsing involved in recovery. Fourth, normal and clinical neuropsychological phenomena will be used to exemplify the consequence of limitations and breakdowns in the processes of recovery and synthesis. Fifth,
several effects in the experimental literature on perception and cognition
will be reinterpreted in terms of characteristics of recovery and of effects
of nonconscious on conscious treatment of an event. Finally some implications of the approach will be briefly touched on. What follows is an
attempt to provide an alternative approach rather than a complete or tight
account of the phenomena in question.
DEFINITIONAL
DISTINCTIONS
The primary criterion for consciousness is phenomenal awareness.
There is a twofold reason for focusing on this, as opposed to other aspects
emphasized by psychologists, such as capacity or behavioral organization
(see, e.g., Shallice, 1978). First, historically, phenomenal experience has
been the raison d’etre for discussions of the concept in the first place
(Strange, 1978; Battista, 1978). Second, whatever other associated properties we may focus on, practical criteria for consciousness, e.g., clinical,
always rely in some sense on awareness, whatever the operational test.
Awareness is taken to be the prerequisite of an ability to acknowledge or
“comment upon” our percepts, thoughts, memories, and actions. However, there is a second definitional criterion which is difficult to separate
from the first. This is the ability to base intentional, categorical action
upon a perceptual (or imaginal) experience. Many aspects of our actions
such as their detailed components or analog aspects can be determined
both at a nonconscious level and by aspects of the environment of which
no conscious acknowledgment need be made (Turvey, 1977a). However,
apart from “reflex” activities, a percept has to be conscious if we are to
voluntarily initiate one action as opposed to another on its basis. In some
sense the criteria1 reason why the representational basis of voluntary
actions has to be conscious is that it is the reason we give if asked why we
did something. In this sense it differs little from the notion of awareness in
that we are concerned with a representation which can be acknowledged
or commented on by the perceiver or actor.
Of course it can be argued that this is an illusion, that such reasons or
comments are not the causes of our behavior, but merely an “account” of
our behavior to conform with, say, cultural beliefs in rationality. It will be
suggested below that phenomenal experience may indeed be thought of as
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
241
an account. Nevertheless, the fact that it precedes action precludes the
argument that it is an account that only occurs after the event. Thus we
can choose or set up what aspect of an event is to control the categories of
our behavior, as opposed to those aspects which merely affect the parameters of the act. In a task or experiment a person can decide to base
categorical responses on whether two strings of letters are visually the
same or different and maintain consistency (be correct), quite apart from
the fact that he or she is affected by (e.g., in the latency of response)
whether or not the stimuli are words.
It should be made clear that this paper is not in the main addressing that
sense of the term consciousness whereby one might ask whether someone
is or is not conscious. Nor is it (at least not primarily) states of consciousness which are to be discussed. Rather, consciousness is addressed in the
sense of whether or not someone is aware of something. It is the intensional aspect of consciousness which is of concern. Phenomenologists,
from Brentano (1874) and Husserl(l929) onward, have held that this is the
fundamental property of consciousness, that consciousness always has an
object, that we are always conscious of something. What is at issue is
what it means to be conscious of something. A central proposal of this
paper is that to be conscious ofx is to be understood as being conscious of
x as a proper segment of space and/or time, i.e., as a primitive at the
relevant level of perceptual organization.
The following discussion will not address the function(s) of consciousness. Nor will it address its nature, except insofar as it is an attempt to
provide one form of process-oriented approach to perceptual phenomena.
OUTLINE
OF THE APPROACH
At this point the basic constructs of the proposed approach and how
they relate will be introduced. In order to clarify the pragmatic force of
the approach, let us return to the starting point. The strongest version of
theory generated by the Identity Assumption is that the representations
produced by perceptual processes and those constituting conscious experience are one and the same entity. This is captured in Fig. 1.
A weaker form of this view, which permits selective attention to different perceptual descriptions and which is plausibly what is actually assumed by many information processing accounts, is to treat conscious
experience as the product of an active attentional process. But this still
treats the information addressed by such a process as identical with that
utilized by subsequent perceptual recoding systems. This is represented
in Fig. 2.
This is essentially the position which the experiments in the preceding
paper have been taken to challenge. It is possible to prevent awareness of
particular perceptual descriptions and the ability to base categorical behavior on them without preventing their unconscious effects on behavior.
242
ANTHONY
J. MARCEL
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
Conscious
PROCESSES
243
Experience
Sensory
Array
(--CM)-----
FIG. 2. Weak version of the Identity Assumption-where conscious experience results
from a separate process, but its representations are those produced and used in perceptual processing. (Here and in following figures, solid lines represent automatic processes,
dashed lines represent intentional/selective processes.)
Further, Experiment 5 demonstrated that it is possible to enhance such
unconscious effects without changing the availability to consciousness of
the cause of the effects. For these types of reasons it is suggested as a
useful first step to separate functionally those representations automatically yielded by and utilized by perceptual analyses from those of which
we are conscious.
What might be meant by such a separation? Clearly, phenomenal experience is not merely a separate copy of nonconscious representations; if it
were so, one could not affect them differentially. Nor can what is conscious be merely a separate subset of what is nonconscious, since much
evidence (some of which is reviewed in the section below on Differences)
suggests that the conscious representation of something is often qualitatively different from its nonconscious representation. The essence of the
present view is that phenomenal percepts depend on nonconscious sensory analysis (or activity in such mechanisms), but also on certain further
operations which render the form of the phenomenal representation qualitatively different. It is from these further operations that the intensional
aspect of phenomenal experience derives. Thus, nonconscious sensory
analysis registers what impinges on external and internal receptors as a
nondisjointed flow (it does not segment into events, objects, episodes); it
codes all aspects of what impinges at every level and in every code with
which the organism is equipped; within each such representational domain what impinges is represented in all possible articulations. Phenomenal experience consists in the imposition of a particular segmentation
and structure on what is otherwise unsegmented (i.e., nonintensional) and
the imposition of a particular interpretation on what otherwise consists of
multiple interpretations. Let us now specify the principal components of
what is being proposed.
244
ANTHONY
J. MARCEL
1. All sensory data impinging however briefly upon receptors sensitive
to them is analyzed, transformed, and redescribed, automatically and
quite independently of consciousness, from its source form into every
other representational form that the organism is capable of representing,
whether by nature or by acquisition. This process of redescription will
proceed to the highest and most abstract levels within the organism.
Further, within every domain of redescription, wherever more than one
parsing is possible of the data presented to it, all possible parsings will be
carried out and represented. The structure of systems which automatically process sensory input is distributed as well as hierarchical. Some
systems depend on the outputs of others, e.g., lexical analysis of alphabetic print depends on a representation in terms of strokes and letters.
However, systems such as form and color operate independently. Good
evidence for such distributed processing comes from neuropsychological
dissociations (Ratcliff & Cowey, 1979).
2. An important functional distinction will now be made concerning the
products of automatic processes of sensory analysis. Each analytic or
redescriptive process which transmits and transforms sensory data obviously has an output, which is a particular data description. Two aspects of
each such output can be distinguished. There is that aspect which, due to
the normal fact that the sensory environment is constantly changing, is
dynamic and impermanent. This will be termed the result. An aspect can
also be postulated which is the temporally extended trace of an output and
which can be retrieved. This will be termed the record. While two terms
are used and while the distinction is made diagrammatically explicit in
Fig. 3, it should be emphasized that we are distinguishing between two
aspects of a single entity, differentiated by the uses to which they are put.
Results have several different automatic effects and roles at a nonconscious level. First, they provide data which is immediately fed to or used
by all other processors which can take such a representational code as
input for further transformations. Second, nonconsciously the outputs of
specialist processors, alone, support a large amount of our behavior, especially those aspects emphasized by Turvey (1977a), i.e., both the basic
context of actions (e.g., postural adjustments) and their detailed realizations, as well as our orientation in space. Third, the results of coding
processes will be in terms of patterns of activation which will in turn
activate structurally related and associated representations in their respective representational domains. (That is, they produce priming.)
Fourth, all levels and types of description will suggest candidate perceptual hypotheses. That is, data descriptions activate canonical representations of permanently stored perceptual and conceptual categories,
which specify features needed to verify them.
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
245
These results of sensory analyses in their role of suggesting hypotheses
form one of the bases of what may become conscious. However, this is
insufficient. Each analytic and redescriptive process also yields a record
of its output, and it is such records which provide the basis of the form
and content of conscious percepts. The record of the output of some
sensory analysis may be inaccessible to consciousness while the result
itself may be fully effective in influencing behavior and quite accessible to
other nonconscious redescriptive systems. This distinction is a functional
one. No claim is being made that results and records are separate entities.
However, in a speculative vein, since records are viewed as supporting
retrieval, some degree of mnemonic persistence would probably benefit
from separation from machinery that is in constant flow. These distinctions are represented in Fig. 3. Both results and records are nonconscious. What relates records to conscious percepts is indicated in Paragraph 4 below.
3. Perceptual hypotheses are themselves unconscious. They are canonical representations of perceptual and conceptual categories at various
levels and constitute the total set of those categories and distinctions that
we can consciously perceive. A perceptual hypothesis is here conceived
of as a structural description which specifies its criteria1 features (plus
noncriterial features) and their relationships. The values of those features
is left open to particular instantiations which vary from occasion to occasion (episodic). Perceptual hypotheses are activated automatically if any
of their criteria1 features are matched by descriptions yielded at the next
lower level of sensory analysis, or if their own category is yielded at the
equivalent level of sensory analysis. Perceptual hypotheses, in representing our knowledge of the perceivable world, are analogous to
Minsky’s (1975) concept of “frames” and to Rumelhart’s (1978) discussion of “schemata.” But note that here they are not the mechanism of
perceptual analysis, but that of conscious percepts. They will be constrained by constructs available at any one time and in any one culture to
an individual. This is one way in which what an individual is capable of
being aware of, and the way in which he or she is aware of it, is moulded.
Indeed, since perceptual hypotheses are the basis of conscious percepts,
they are the basis of the particular intensionality of our experience.
However, neither the records of sensory analysis nor perceptual
hypotheses are sufficient for conscious percepts. A conscious percept is
obtained by a constructive
sensory source.
act of fitting
a perceptual
hypothesis
to its
Phenomenal experience has to have some form, and hypotheses will
remain unconscious unless they bear some relation to some description of
.:
Hypotheses
Type A
Hypotheses
Type B
FIG. 3. Illustration of the proposed distinctions within automatic perceptual processing. Processes which analyze and transform information produce a resulting representation and a record of it. Results (a) serve as input to other processors, (b) automatically modulate mechanisms relying on activation (priming, motor adjustments), (c) produce perceptual hypotheses. Perceptual hypotheses guide recovery of
information from records, which serves to verify alternative hypotheses. Conscious percepts normally require synthesis of verified information.
sensory ....”
Array “. . .
g
r
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
247
sensory data.’ While we may not be sure how to interpret what we see or
hear, we must at least be conscious of something. In addition, since the
outputs of sensory processes can suggest alternative perceptual hypotheses, the data from records will serve to disambiguate. This is also necessary since we can only be conscious of one perceptual hypothesis within a
particular domain at a time. For the moment we can conceive of the
process as one of verification.
The relationship of hypotheses and records to conscious percepts can
be summarized as follows.
(a) We choose at what level to be conscious.
(b) This involves the parallel testing of a subset of the activated perceptual hypotheses at the chosen level against appropriate records.
(c) One hypothesis will be selected as a result of this process partly on
grounds pertaining to the hypotheses (e.g., expectancy, frequency),
partly on grounds pertaining to the records, i.e., what will account
for most data in the record, and partly on grounds pertaining to
both, i.e., highest possible level of description.
(d) At any moment what we are aware of is that hypothesis instantiated
by the fit of data in the relevant record.
(e) We are unaware of hypotheses tested but not verified or insufficiently verified.
(f) Equally, both records and results remain unconscious; it is only
records recovered by a unitizing hypothesis that are conscious.
(g) We are unaware of the processes by which hypotheses are chosen
for testing and by which they are tested.
(g) Consequent upon the selection of one perceptual hypothesis, other
competing hypotheses will receive inhibition. Note that inhibition is
a concomitant of selection, rather than being the mechanism of
selection (as, for example, in Shallice’s 1978 model).
4. If a hypothesis is to be matched against the relevant sensory data,
then that data has to be retrieved. This process will be termed Recovery of
information. Information is recovered from the records of sensory
analysis. This notion captures the idea that conscious experiences is a late
r If a perceptual hypothesis needs to be matched against the records of sensory analysis, it
might seem that we could not be conscious of dreams, images, or hallucinations which have
no sensory source. First, we can suppose that the activation of hypotheses can in turn
activate, in top-down manner, mechanisms of sensory analysis not already driven by the
environment (Neisser, 1970). Second, it is of interest that the specific sensory aspects of
internally produced images are phenomenologically
not well particularized and are unstable.
Thus records at the appropriate level (the criteria1 features of a perceptual hypothesis) would
be produced and recoverable, but not those at lower or more analytic levels.
248
ANTHONY
J. MARCEL
stage and that the ability to consider, or base actions on, some aspects
rather than others of events is selective. Thus when one’s consciousness
is focused on either the meaning or the sound of an utterance, perceptual
hypotheses are produced equally regarding both but recovery will be
primarily of one.
The process of recovery will be dealt with in greater detail later on, but
two points should be mentioned now. First, perceptual hypotheses, which
form the contents of consciousness will be in discrete terms and in terms
of “objects” or “properties.” They are segmentations of the world. That
is, they are accounts of the world or of experience in terms of structural
descriptions (Palmer, 1977), and at the level of the object domain (Clowes,
1971). There is little to suggest that this is a pervasive aspect of perceptual
analysis itself. Therefore a central aspect of the process of recovery is that
it parses the contents of records. Further, since perceptual hypotheses
may account for only part of the information in the appropriate record,
recovery serves to articulate the record into a structured subset of the
possibilities which are simultaneously present. This structured subset is
what becomes conscious. That the process of becoming conscious is one
of organization will be central to our understanding of masking discussed
below.
Second, while perceptual analysis proceeds in the direction of more
derived descriptions, recovery works in the opposite direction, it being
harder to recover those descriptive records which are more analytic or
closer to the proximal stimulus. The view advanced here is that phenomenal experience is an attempt to make sense of as much data as possible
at the highest or most functionally useful level possible. Learning and
development “push” our consciousness to ever higher levels. Thus as we
learn to interpret the significance of a cue or combination of cues, we are
aware of that significance instead of and before we are aware of the cues.
The present point of this is that the nearer the attended level of representation is to that level at which we normally operate functionally, the
easier it is to recover. That is, deciding what word was spoken is easier
than deciding what phoneme was spoken, because the former is nearer to
the semantic and pragmatic levels of representation. In addition, although
automatic representation of a higher level is only achieved after that of a
lower level, by being derived from it, nonetheless since this happens
before any recovery operation, recovery of lower level information is
necessarily influenced by any higher levels of representation that have
been achieved.
5. A further essential aspect of the process of recovery necessary to
instantiate perceptual hypotheses is that of the Synthesis of the records of
more than one specialist processor. Integration of different types of information is necessary for several reasons. First, as noted above, if the
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
249
output of lexical analysis gives rise to a hypothesis at the lexical level,
confirmation cannot be by recovery of a lexical record alone; at least the
record of the sensory form or source of such a description is necessary.
Otherwise one could not say “I saw such a word” as opposed to “I heard
such a word.” Thus to have a particular sensory experience (visual, auditory, etc.) requires records of at least two levels of analysis. Second, in
the visual domain at least both form and location are required to articulate
the visual field in terms of objects. Third, without integration of different
types of information, our percepts would not have the phenomenal unity
that they do usually have. That is, normally we are not aware separately
of a certain color, a particular shape, a texture, and the idea of a table,
each located at a particular point in space, but of one thing: “a-table-ofa-certain-color-and-wood-oriented-in-a-particular-way-at-a-certain-location-and-distance.” Recovery alone of feature values will suffice for
awareness, but will not be sufficient for normal unitary percepts. If
synthesis does not operate normally and at the appropriate level, a person
will be quite aware but their percepts will lack appropriate unity. (It is
proposed in the clinical section below, that cases of disjunctive agnosia
exemplify such an impairment.)
Synthesis could be carried out on two (logically distinct) bases. (a)
Individual properties of the optic array are represented in records, at each
level of description, with reference to a spatial map. (b) Perceptual
hypotheses not only specify their features (whose values are instantiated
by data in the records), but also specify the structural relations between
all their features. When a hypothesis is verified the structural relations in
the hypothesis permit the appropriate binding together of the features in a
structure. Both the features of objects and the objects in a scene are
plausibly related one to another by structural descriptions at each level.
The feature values (e.g., color, texture) are plausibly related to each
feature by a spatial map of their origin in the array. Thus, the instantiation
of a perceptual hypothesis involves the use of hypotheses as frameworks
to both recover and synthesize the information specified by that hypothesis.
6. A synthesized percept is obviously not unitary to the extent of the
simultaneous conscious embodiment of all aspects or levels of information whose representations have been achieved by perceptual processors.
This is in contrast to Allport’s (1977) claim that it is. We can and do
retrieve the records of only some of the processors. We choose to attend
to alternative levels or aspects of events.
7. The processes of Recovery and Synthesis are related to three aspects
of consciousness: Awareness, Unity, and Selectivity. Recovery of information is necessary and sufficient for awareness. Synthesis, which depends on recovery, is necessary for attribution of a percept to a source
250
ANTHONY
J. MARCEL
and for the unitary nature of the percept. Selective recovery of information is what underlies both the level or aspect of our awareness and our
ability to set up or choose which aspect of what we perceive will control
categories of behavior. Thus selective recovery enables us to decide (a)
whether or not two letters are the same graphically (AA) or graphemically
(Aa) (level of awareness), or (b) whether or not two letters are the same
size versus the same color (aspect of awareness). This relates to the
logical requirements of tasks or the categorical status of an act rather than
its analog aspects (speed and manner of execution).
8. The foregoing has covered the bones of the present proposal and is
illustrated in its barest information-processing form in Fig. 4.
The view presented here is in contrast to that of authors such as
Deutsch and Deutsch (1963), Dixon (1971), or Shallice (1972, 1978) who
have treated the transfer from preconscious to conscious perception
primarily as a matter of strength of activation. This is not only because the
results of Experiment 5 of the preceding paper seem to militate against
strength as a sufficient criterion. The other main reasons are that it will
permit (a) treatment of qualitative differences between conscious and
nonconscious representations, and (b) a principled treatment of otherwise
paradoxical perceptual phenomena. The present emphasis on qualitative
differences also differentiates this from the views of Lashley (1958), Miller (1962), and Mandler (1975). These authors are careful to say that
mental processes are not part of consciousness, but that the results of
mental processes are what are conscious. Implicitly, according to all the
above-mentioned authors, all that can be represented consciously is represented in the same fashion nonconsciously, and therefore becoming
conscious of something is merely like opening a door to, or it being
pushed by, one aspiring entrant. According to the present view phenomenal experience is an attempt to make sense of as much data as possible
at the most useful level, according to culturally given presuppositions,
and therefore it involves a further, inferential step.
DIFFERENCES BETWEEN CONSCIOUS AND UNCONSCIOUS
PROCESSES AND REPRESENTATIONS
In the outline of the approach so far, the main difference mentioned
between conscious and unconscious processes has been automaticity versus selective intentionality. The distinctions between the two classes of
perceptual processes can be further elaborated. However, the importance
of discussing such distinctions is not merely elaborative. Certain types of
putative differences have implications for the nature of the processes of
recovery and synthesis. This is why this section precedes a more detailed
exposition of those processes.
FIG. 4. Summary of proposed position. This represents (a) the distinction
between automatically
yielded representations
and those of
which we are conscious, and (b) the course from (i) nonconscious representations
to (ii) perceptual hypotheses to (iii) recovery to (iv) verification of hypotheses to (v) synthesis of information in a conscious percept.
-M
CT:
I
252
ANTHONY
J. MARCEL
Capacity
While unconscious automatic processes are not bound by capacity,
those associated with consciousness are. Recovery of different types of
information and decisions based on different types of information are
restricted in their temporal parallelism. Thus Treisman’s recent work on
the limited capacity of different perceptual decisions (Treisman, Sykes, &
Gelade, 1977) and Marcel’s earlier findings on within-channel seriality
(Marcel, 1970) refer to limitations on recovery rather than initial processing. In both these studies subjects were relatively unlimited in dealing
simultaneously with stimulus dimensions per se (color or shape) but were
limited in their ability to deal with conjunctions of dimensional values (a
red circle or a green square). A further indication of representational
capacity is provided by the present author’s work on polysemous words
(Marcel, 1980). Apparently more than one interpretation of an event in
any one domain can be represented simultaneously nonconsciously but
only one interpretation at a time can be represented consciously. In the
lexical-semantic case investigated by Marcel this applied to the meanings
of words such as PALM. However, this principle also applies to more
perceptual domains. Thus if a Necker cube is seen as a cube then consciously it can only be seen in one orientation at a time. However, prior to
that stage of perceptual synthesis not only can it be represented in both
orientations simultaneously, but each corner can be represented as both
concave and convex. This can also be seen to apply to lexical segmentations of auditory streams. The acoustic pattern phonetically represented
as [wetta] will be automatically segmented by a British listener into both
“waiter” and “way to” at an unconscious level, but only one lexical
structuring of it can be represented at one time consciously.
This difference in capacity applies not only to representation of a single
event but to the processing and representation of simultaneous events.
While we can only recover information from one speech stream at a time,
the experiments of Lewis (1970) and Corteen and Wood (1972) suggest
that simultaneous analysis of different streams proceeds automatically to
at least a lexical level. In the visual case evidence has been provided by
Bradshaw (1974), Underwood (1976) and Willows and Mackinnon (1973)
that words which cannot be reported are analyzed simultaneously with
attended material. Indeed in order to produce the very initial determination of eye movements as reported by Yarbus (1968) and Mackworth and
Bruner (1970), different segments of the visual field must be simultaneously analyzed to high levels of interpretation prior to allocation of
attention.
Intention
and Znhibition
A second distinction, which is often theoretically coextensive with the
terms conscious and nonconscious, is between intentional and noninten-
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
253
tional processes. The problem is that while commonsense theorizing is
based on the distinction, what is needed is an operational criterion and
qualitative differences consequent upon it. However, it is possible to find
phenomena which fulfill these criteria. One such is lexical/semantic
priming. It would appear that trial-to-trial facilitation can be determined
both by an intentional process and unintentionally. Tweedy, Lapinski,
and Schvaneveldt (1977) have shown that the extent of facilitative semantic effects by prior context in lexical decision are dependent on the proportion of successive pairs of semantically related words. On the other
hand, Fischler (1977) has shown that associative effects occur in this
situation without any prior expectation of an associated pair. The crucial
difference is suggested by an ingenious experiment by Neely (1977). His
results strongly indicate two types of contextual effects consistent with
the theory delineated by Posner and Snyder (1975). The first is a fast,
automatic inhibitionless spread of activation to semantically associated
lexical entries, which proceeds independently of a subject’s expectancies,
intention, or control. The second is a slow effect which depends upon the
subject’s intention and conscious awareness and which can be focused at
will on arbitrarily defined sets of lexical entries but which involves the
generation of inhibition to those entries unrelated to those on which it is
focused. Interestingly enough this formulation, especially the dependence
of inhibition on consciousness exactly fits Marcel’s (1980) findings with
polysemous words. When PALM was masked to prevent consciousness it
facilitated lexical classification of WRIST whether it was preceded by
HAND or TREE. But when PALM was not masked and the subject was
conscious of it, it had an inhibitory effect on classification of WRIST
when preceded by TREE. Indeed Posner and Snyder’s theory of conscious inhibition supported by Neely’s data would account for one aspect
of the results in the present Experiment 3 (Marcel, 1983). When subjects
were aware of the word stimulus there was a greater development of
interference relative to facilitative effects with longer word-color intervals than when they were not aware of it.
While the foregoing has attempted to instantiate the association of intention and nonintention with conscious and unconscious, it is plain that
what is initiated consciously is realized in large part nonconsciously. Thus
while the present position (Posner & Klein, 1973; Marcel, 1980) argues
that inhibition is a concomitant of conscious intention, it operates at a
nonconscious level. In the case of action, inhibition is not only generated
in the selection of intended, voluntary actions, but operates in the modulation of that postural and motor control which, though serving as a
supportive background, is not the focus of intended or consciously regulated behavior (Turvey, 1977a). Carr and Bacharach (1976) have pointed
to the counterpart of this in perceptual mechanisms. Thus there are two
sources of both activation and inhibition. Firstly, in perceptual analysis
254
ANTHONY
J. MARCEL
prior to consciousness several entries may be activated either by separate
sources or by association without inhibition. At the stage of consciousness or “focal attention” unsuccessful competitors for spatial attention
and for the representation of any one spatial source will receive inhibition.
Secondly, in postconscious intentional processes such as search of lexical, semantic, or other representational memory systems, activation entailed by the focusing of conscious attention itself entails inhibition of
nonattended categories. In essence, the line taken here is to associate
inhibition only with those processes whose issue is constrained to be
selective. Selective bounds are put on actions to the extent of their coherence and biomechanical constraints. They are put on conscious perception to the extent of its coherence and limited capacity.
To illustrate the boundaries on activation and inhibition, consider the
nature of one type of error in the consciously initiated activation process
of lexical selection in speech production. Let us suppose that lexical
selection is driven on the basis of semantic and syntactic specification and
results in phonological output. If we ignore those speech errors which
result from the context of the rest of the utterance (e.g., anticipations,
perseverations, transpositions) one of the predominant classes of error is
that of blends of two potential candidates. Examples from Fromkin (1973,
Appendix U) include “momentaneous”
(instantaneous/momentary),
“herrible” (terrible/horrible), “smever” (smart/clever). In some sense
utterance of a word is selective or categorical by force of structural
constraints-we cannot utter two words at the same time. Now to a
certain extent the existence of blend errors of the type exemplified is an
example of failure of inhibition. But then the intended activation is of
something which satisfies the semantic, syntactic, and pragmatic requirements, and in the cases listed in Fromkin’s corpus both of the lexical
entries leading to the blend in all cases satisfy those requirements. What
one does not obtain in blend errors is (a) the intrusion of semantically
unrelated words, or (b) substitution of segments related only phonologically (unaffected by surrounding context). Indeed the only case where
these two other sources of intrusion appear to contribute to blends is in
jargon aphasia. When they do it appears either due to lack of inhibition or
relaxation of inhibition (Butterworth, 1979). Thus, the existence of blends
due to one source (semantic activation) and the normal nonexistence of
other sources (inhibition of lexical segments related only phonologically)
is a demonstration of categorically defined activation and categorically
defined inhibition.
Representational
Categories
The preceding discussion of associative activation and concomitant inhibition has raised the question of the boundaries of such processes. It is
thus appropriate to discuss at this point another distinction between con-
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
255
scious and unconscious processes-the form of the knowledge and representations upon which such processes operate, i.e., discriminative
categories and their dimensions of coding and association. In Polanyi’s
(1964, 1966) terms, the tacit and explicit (i.e., nonconscious and conscious) descriptions of an event may be quite different. Turvey (1974) has
already made this point and presented some of the evidence which follows. Wickens (1970) has argued that release from proactive interference
(PI) in short-term memory (improvement in recall performance following
continuous deterioration) reveals psychological categories by dint of differences in coding between classes. In an interesting set of studies, Turvey used sets of words chosen from the different poles of each dimension
of the Semantic Differential (Turvey & Fertig, 1970; Turvey, Fertig, &
Kravetz, 1969). Shifting across dimensions within the same polarity did
not produce release from PI; however, shifting polarity either within or
between dimensions produced highly significant release from PI. Yet if
people are presented with words differing in their polarity they cannot
intentionally discriminate or sort them. In other words a distinction is
being made tacitly or unconsciously (as revealed by release from PI)
which is not being made consciously.
The opposite of this is to be found with syntactic form class of words.
While one can quite easily distinguish consciously between nouns and
verbs, Wickens (1970) has found that a shift from one to the other has no
effect in release from PI. Again, a shift from concrete to abstract and low
to high imageability words fails to have any effect on PI (Wickens &
Engle, 1970) while imageability is found to be an important variable in
intentional learning and memory (Paivio, 1969). This difference has been
echoed in the contrast between reportable word perception and unconscious registration. Marcel and Patterson (1976, 1978) found that imageability interacts with hemifield when subjects are asked to report tachistoscopically presented words (i.e., low imageability words are much harder
to report at the same target-mask SOAs in the left visual field) and that
imageability affects lexical decision time. However, entirely different implications were yielded in an experiment where low- and high-imageability
words were masked to prevent awareness and used as primes for subsequent lexical decision words, both within and across hemilields. When
thus masked and their effects measured as in Experiment 4 of the preceding paper (Marcel, 1983), there appeared to be no difference either in the
representation or accessibility of high- and low-imageability words in
either hemisphere nor in their efficacy in semantic priming of associates,
within or across cerebral hemispheres. Something similar to the studies of
release from PI can be seen in the clinical field. Warrington (1975) found
that certain agnosic patients could neither read nor recognize particular
words nor auditorily discriminate them from nonwords. Yet their immedi-
256
ANTHONY
J. MARCEL
ate memory span for these words with auditory presentation was signiticantly better than for the nonwords. A categorical feature affecting behavior (lexical status) was not recoverable consciously as a feature upon
which to base intentional categorical discriminations. A further example
is that in Lewis’s (1970) experiment synonymity of unattended to attended
words affected reaction time to attended words, but associative relations
derived from associative norms did not. Turvey argues from this that
associative norms reflect explicit cognitive dimensions but are not
isomorphic with the structure of tacit knowledge.
In the cases cited so far it appears that certain dimensions are relevant
to the recovery of information about and conscious use of different types
of verbal material, while different dimensions are relevant to their nonconscious representation, processing, and effects. It has long been
claimed that radically different principles apply to conscious and nonconscious processes. A common line is that while the structure of conscious
knowledge is logical, that of nonconscious knowledge is merely associative. This section while aimed at supporting a distinction, is an attempt to
hint at more subtle differences.
More Radical
Differences
in the Codes-Implications
for Recovery
The immediately preceding discussion has suggested some of the differences between conscious and unconscious representations. If these were
only differences in the relative salience or availability of particular characteristics (e.g., syntactic class versus connotative features of words),
then they would be elaborative but not central to the present concerns.
However, certain approaches suggest that the structural languages of
conscious representations are not directly mappable onto those of nonconscious representations, i.e., they are neither commensurate nor
coextensive. If this is the case then either (a) what have been termed the
records of the outputs of perceptual analyses are not mere copies of such
outputs but translations, or (b) the process of recovery of information is
not so much retrieval as the imposition of a qualitatively different structural description. The type of problems which lead to these notions can be
illustrated from the realms of speech perception and visual perception.
It is often assumed that the perception of speech proceeds from an
acoustic description of the signal to an intervening phonetic description
and thence to a syllabic or lexical description. However, certain lines of
evidence from simulation suggest that the code used by lexical categorization for auditory word recognition could be in directly acoustic rather
than phonemic terms (Warren, 1976; Klatt, 1979; Marcus, 1979). Another
cogent argument (Oden & Massaro, 1978; Massaro & Oden, 1980) is that
the smallest units capable of reliable isolated recognition are syllables, but
again that syllables use acoustic and not phonemic information. Yet it is
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
257
arguable that the analytical limit of our conscious experience of speech
sounds (locutionary awareness) is restricted to phonological descriptions;
and that even our ability to experience phones or phonemes rather than
syllables is a special acquisition (Morais, Cary, Alegria, & Bertelson,
1979; Marcel, 1979). We just cannot auditorily experience, directly and
selectively, acoustic features or segments, but have to infer them. Additionally while phonological descriptions of the speech signal may be considered abstractions of acoustic descriptions, phonological and acoustic
segments of the speech signal are not coextensive. Not only do the informative acoustic segments extend over wider ranges than phonemes
(Liberman, 1970, Fig. 5), but, for example in initial consonant clusters of
stop+lr/, one set of acoustic cues indicates a single phoneme while others
indicate two (different) phonemes (Kornfeld, 1977). As Studdert-Kennedy
(1981) admits, the enterprise of trying to find a mechanism for translating
the acoustic signal into phonetic segments has been given up. In fact, the
possibility exists (see Marcel, 1979) that phonemic descriptions are an
invention which have a phenomenological but not a processing reality, at
least in sensory analysis. The problem this creates is as follows. Suppose
we are listening to speech and attending to the phonemes uttered. How
could we directly recover phonemes from the records of any level of
speech analysis if speech analysis does not itself represent them? They
are not explicitly represented in syllables, and acoustic segments are not
coextensive with them and are ambiguous. The potential solution proffered by Wickelgren (1976), that phonemes could be directly derived from
context-sensitive allophones, has been discredited by Massaro and Oden
(1980, p. 135). The fact that we can in fact consciously attend to phonemes
seems then to demand that the process of recovery is rather more complicated than simple matching.
In vision, the perception of depth, motion, and shape over time provides a similar problem. Our conscious experience is in terms of separate
rigid bodies describable in terms of static metric Euclidean geometry. Yet
the ligural invariances of the environment with respect to viewpoint,
depth, and motion are more economically described by nonmetric projective geometry (Johansson, 1974; Shaw & Pittenger, 1977). While projective geometry is theoretically mappable onto Euclidean geometry, enormous problems have remained throughout attempts over the last century
to account in Euclidean terms for visual processing itself of space. Thus
Western conventions of pictorial representation, especially those traceable to the Renaissance (e.g., the use of linear perspective to represent a
single visual moment from a single retinotopic viewpoint), can be seen as
making explicit the analytic limits of our visual awareness (in the same
sense as locutionary awareness) rather than representing the language of
the visual system. This is an extremely important point which will be
258
ANTHONY
J. MARCEL
taken up in the next section. First it allows us to specify the appropriate
domains of discourse for on the one hand physical% and biological descriptive metrics and on the other hand cognitive descriptive metrics.
Second, if valid, it sets severe limitations on the ability of psychophysics,
which relies on phenomenally based judgments, to reveal the nature of
sensory systems. However, the important point here is that if the geometry of visual processing is qualitatively different from that of phenomenal
visual experience, recovery cannot be a simple matching process.
THE PROCESSOFRECOVERY
Recovery
and Verification
If the languages of conscious experience are often categorically different from those yielded by sensory analysis, as suggested in the last section, how are the two related in the process of recovery and verification?
In the section outlining the current approach, verification was described
as an attempt to relate perceptual hypotheses to their counterparts in the
records. As such it could be viewed as a simple parallel search and comparison process (i.e., for hypotheses of Level/Type A, “Is Hypothesis Aj
contained in Record Ai
. “?” where i . . . IZ defines a spatio-temporal
extent). However, comparisons are only valid if what are compared are in
the same or compatible codes. Since the previous section suggests that it
is probable that the codes are not directly mappable, the conception of the
verification and recovery process needs modification.
One approach to this problem is to take seriously the notion of conscious experience as an account, an attempt to make sense of as much
data as possible in terms consonant with our beliefs. This view has been
applied to various areas of experience-perceptual
illusions (Gregory,
1970), emotional experience (Schachter, 1973, experience and use of
language (Campbell, 1979). In applying this view to the process of verifying perceptual hypotheses against recovered information, the most appropriate metaphor for a mechanism is Piaget’s notion of Assimilation
(Piaget & Inhelder, 1969), whereby the environment is understood in
terms of our internal schemata. Moore and Newell (1974) present one of
the few attempts to realize computationally a mechanism of assimilation
which does not finesse or trivialize the issue of incongruent languages.
The program (MERLIN) achieves an interpretation of the external language by trying to view the external language in various ways, noting
what entities have been attempted but failed to match. “Supposing”
equivalence between such entities, it attempts to see if other entities will
be mappable with relaxed equivalence criteria. It attempts to impose ever
more abstract schemata onto entities in the external language, each time
withholding equivalence criteria for certain entities or segments of the
259
CONSCIOUS EXPERIENCE AND PERCEPTUAL PROCESSES
/
/I
/’
//
/
/
/’
,/
/R _ - Giii$r-&,
\
‘I:,
1
FIG. 5. The process of verifying perceptual hypotheses against recovered information.
Where codes are not compatible, verification is represented as assimilation by inferential
problem solving.
external language, until a satisfactory range of the external data can be
accounted for. The equivalences postulated by the accepted account but
not yet strictly verified will then be accepted. The mechanism described
by Moore and Newell (1974) seems an appropriate way of viewing the
process of accessing and handling the records of sensory analysis for the
purpose of synthesis into a unitary percept. Note that the question asked
in the verification process is now in terms of criterial matches of the form
“Can anything be found in Record Ai
,, which can be taken to support
hypothesis a,
X, and if so which hypothesis can be best supported?”
(preference being given to higher level or more encompassing hypotheses). This is captured rather crudely in Fig. 5.
Constraints on Recovery
Firstly some records are usually always inaccessible to consciousness,
at least in the absence of some special training procedure. For example it
is arguable that we can never base categorical actions on the most
peripheral processes contributing to form analysis, nor may it be possible
to attend directly to acoustic features of speech. Secondly such records
have very limited lives. Their duration may differ according to their type
or other factors. Thirdly it is easier and possibly quicker to recover information from a higher level than from a lower level. As we learn to interpret the significance of a cue or a combination of cues, we are aware of
that significance instead of and before we are aware of the cues. We hear
words not phonemes and their coarticulation. To use an example of
260
ANTHONY
J. MARCEL
Polanyi’s (1966), when we are exploring the interior of a hole with a stick
we are aware of the shape of the hole not the pressures of the stick on our
hand, although the former is mediated by the latter. In a general sense the
level of our consciousness is pushed to the level of whatever representations we learn, and the nearer the level of the attended representation to the level at which we operate functionally, the easier it is to
recover. This relation between ease of recovery and level of representation goes toward an account not only of the order of dropout of the three
types of decision in Experiment 1 of the preceding experimental paper but
also of phenomena such as the Word Superiority Effect and the influence
of higher order factors in phoneme monitoring. This will be discussed
below in the section on reinterpretation of such phenomena.
Fourthly, where multiple representations are yielded in any one domain,
for a particular time or spatial location, then only one can be recovered at a
time. In audition two words may have been uttered by different speakers,
or one acoustic pattern may be structured into different phonemic or
morphemic segments (e.g., [werta] can become either “waiter” or “way
to” for a British listener). In vision different patterns may be projected on
to spatially equivalent parts of the two retinae or one pattern can be
structured in two ways (e.g., a Necker cube). Examples at the semantic
level are polysemous words (race) and at the syntactic level polysyntactic
phrases (they are flying planes). In any domain nonconscious processes
will yield all possible representations. But for any domain only one value
can be recovered at a time for a particular point in the appropriate multidimensional space. The source of this boundary on recovery is in the
contrast between the relatively unlimited processing and representational
capacities of nonconscious systems and the relatively limited processing
and representational capacities of conscious systems.
Intimately related to the foregoing is the nature of the codes which are
synthesized. For example suppose one is attempting to recover the letters
of a letter string. If a lexical code is available (i.e., if they form a known
word) then that is more economical than a purely graphemic code. Thus
given the capacities of consciousness, as revealed for example by studies
of short-term or working memory, more graphemic information can be
recovered if the graphemic record can be synthesized or matched with a
lexical record than if not. In general more lower level information can be
recovered the more economical a code it can be synthesized with. Indeed,
a point that is depicted in Fig. 4 is that records of more analytic representations can be addressed via records of representations derived from
them, but not vice versa.
To put this in a slightly different way, those things of which we are
conscious are structural descriptions. While perceptual processors may
work on data which is at the level of nonparsed elements, the perceptual
hypotheses which constitute consciousness and which are matched to
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
261
recovered information are always structural descriptions which correspond to parsings of elements. This idea is essentially the view put forward by Hochberg (1970). Thus the capacity constraints on recovery are
not due to the number of elements that can be handled but due to the fact
that only one structure can be used at a time as a synthesizing principle for
recovery. In this sense perceptual spans reflect the level and economy of
organizations imposable on records.
However, this aspect of recovery carries with it another constraint,
which will be used to explain masking and is more fully explored in that
section below. Since what is used to recover information is a structural
description, what recovery does is to impose a parsing on that which is in
a record. The first aspect of this is segmentation of a record. It is supposed that there is an upper bound on the number of independent segments that can be recovered which are not themselves linked by a
superordinate structural description. The second aspect is that for whatever is treated as a single segment, its constituent elements will not at the
same time be recoverable as discrete entities. Thus, while perceptual
hypotheses, which guide recovery, attempt to account for as much data
(elements) as possible in any spatially or temporally delimited portion of
the record, they will often not account for all data therein. Therefore by
its very nature, the process of recovery gives privilege to certain elements
of the record in attaining conscious status to the detriment of the remaining elements.
Lastly, recovery is bounded by interference. If a result of processing
has not yet been recovered to become a conscious percept, then its replacement in space or time in the record of that process will prevent such
recovery. This is the proper application of Kolers’ (1962) clerk-customer
analogy. In Kolers’ analogy a clerk who is asking a customer certain
questions may be hurried or disturbed by a subsequent customer. But this
limited-capacity, distractible clerk is not a mechanism which analyzes or
transforms information (as in Turvey’s, 1973, use of the analogy) but a
mechanism which recovers information. This again relates directly to
masking which will be discussed below.
CENTRAL
MASKING
AS A STRUCTURAL
ASPECT OF RECOVERY
The purpose of this and the following two sections is to clarify the
present approach by applying it to specific phenomena. This section presents a functional approach to central visual masking in terms of
phenomenological goals. Specifically, empirical characteristics of central
masking will be examined as illustrations of the principle of structural
organization and of other constraints on recovery.
262
Preliminary
ANTHONY
J. MARCEL
Remarks
Masking refers to those situations where the presence of one stimulus
impairs the perception of another stimulus in close temporal and/or spatial
proximity. Although several considerations make it difficult to define the
realm of discourse with regard to specific masking procedures, one broad
distinction will be made. Turvey (1973) distinguished between peripheral
and central visual masking. The former is only obtainable if the two
stimuli are presented to the same eye, it is a function of the energy relations between the stimuli, and it works in both a forward and backward
manner. The latter is obtainable whether the stimuli are presented to the
same or different eyes, it is a function of stimulus-onset asynchrony, it
tends to work in a backward manner (the latter stimulus masks the earlier), and it requires some structural similarity of pattern between the two
stimuli. Energy masking is here assumed to impoverish the information
which automatic preconscious sensory processes use. Thus it precludes
any results being produced by earlier processors of sufficient definition to
be further transformed. In the case of central pattern masking, it will be
assumed that preconscious processing is unaffected and that the masking
interaction interferes with some aspect of the achievement of a conscious
percept.
An Approach
to Central Masking
Before describing the proposed conception it is appropriate to present
data which strain the credibility of current approaches and which will be
returned to as exemplifying the principles of the proposal. Current approaches assume that central masking interferes with or terminates processing at the level of visual analysis. Various kinds of evidence already
suggest otherwise. These types of data need integtating into an approach
to masking. Firstly there is the work by Dember and Purcell (1967), which
has been replicated and extended by Kristofferson, Galloway, and Hanson (1979). In these studies identification of a target letter is impaired
when followed at certain SOAs by a mask, but when the mask itself is
masked by metacontrast by a subsequent stimulus then identification of
the target is recovered. (Note that here we are not talking about recovery
under peripheral or brightness masking as in Dember, Schwartz, and
Kocak’s (1978) or Turvey’s (1973) studies.) Either target-mask interactions and stimulus identification do not occur until much later than
single-mask experiments would lead us to expect, or masking does not
interfere with or terminate visual analysis itself. A second phenomenon
which is consistent with the latter idea is that reported by Schultz and
Eriksen (1977). In this case the target was one of four digits, discriminable
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
263
and recognizable but with incomplete contour strokes. Forward and
backward masks were (a) “enhancing’‘-i.e.,
included elements which
replaced the segments missing in the appropriate target, (b) “confusing” -i.e., included elements missing from another target, thereby decreasing discriminability, or (c) “noise’‘-i.e.,
containing random elements. As IS1 decreased, relative to a no-masking condition the enhancement mask facilitated target recognition, while the other two masks
impaired recognition. Plainly the masks cannot have been interfering with
the visual analysis of the targets.
Evidence of a different, quite dramatic kind has been provided by
Jacobson, in studies which went a long way to meeting methodological
objections to his earlier (1973) experiments (1974; Jacobson & Rhinelander, 1978). He showed that while longer SOAs are needed to preserve
verbal identification of masked words as the masks become graphically,
graphemically, and lexically more similar to the targets, much shorter
SOAs can be tolerated when the mask is a word semantically associated
with the target than when it is not. If the mask had impaired or preempted
visual analysis itself, which must precede lexical representation, there
could be no lexical-semantic interactions between target and mask.
Of course the experiments presented in the preceding paper (Marcel,
1983) provide perhaps the most substantial evidence. Words masked so
that they cannot even be detected show effects of lexical characteristics
derived from their visual form on other words.
The present approach to central masking addresses itself to two questions. (1) What effect does masking have? (2) What principles govern
whether two stimuli will interact to produce masking or some other effect? As will be seen, these two questions are not independent. However,
regarding the first, one effect of masking under special conditions is to
interfere with recovery of certain spatial, temporal, or other physical
elements of one or more of the records of the results of sensory processing, as discrete segments. Depending on the nature of target-mask relationships, masking may preclude awareness at all of a certain source of
information, or within that source of information may preclude recovery
of different levels or types of information (location, color, identity). Without segmentable evidence of particular form or of particular location, the
separate existence of an environmental event or aspect cannot be acknowledged or experienced. Less severely, a high-level aspect (identity,
meaning) may be recoverable, but not lower level aspects (location,
form). These lower level characteristics are the more variable (or
“episodic”) aspects of events, and the performance of many perceptual
tasks (reporting from brief visual presentations on the basis of form,
color, or location) depends on the synthesis of recovered information at
high and low levels.
264
ANTHONY
J. MARCEL
This leads to the second question. To approach the principles of interaction which govern what is available to consciousness, it is necessary
to return to the principles of phenomenal experience and of recovery
suggested earlier. Phenomenal experience is intensional, in that what we
perceive are objects and patterns in space, and events and episodes in
time. The aim of recovery and synthesis is to make sense of or account for
as much data as possible in the above terms, i.e., event perception; and
what is used to recover data from records is a structural description, i.e.,
phenomenal experience is a parsing of the records. “Events” are restricted portions of space and time (Cutting, 1981). When objectively
separate events are treated as in the same portion of space/time, only a
single event can be consciously perceived, since only one structural description can be chosen at a time for a particular space/time segment.
With regard to what is recovered, it was suggested that higher level descriptions are chosen if possible (i.e., are more accessible to recovery),
and that economy of available description will govern amount recoverable.
Let us now make these general remarks rather more specific. To begin
with, let us suppose that what governs whether two events will be in an
interactive relationship at all is whether they are treated as being within
the same parsed segment of a structural description or not. An extremely
good illustration of this notion is provided by Bregman and Rudnicky
(1975) in the auditory domain. A portion of a repeatedly cycled signal
stream consisted of the four tones XABX, where X was of lower frequency than A and B and the tones before and after the Xs were lower
still. The listeners’ task, to identify the order of A and B, was made
difftcult by the flanking Xs. But by moving the surrounding tones nearer
to the frequency of the Xs, the Xs were “captured,” leaving A and B in a
subjective stream of their own and making their order easier to identify.
Consider the studies of central visual masking by Dember et al. (1967) and
Kristofferson et al., (1979) described above. When all that is presented is
a target and mask, then provided they are within certain limits of temporal
separation, nothing acts to prevent their being treated as belonging to the
same structural unit. But if a further event occurs, it is possible to induce
an alternative temporal structuring whereby the mask and the additional
event are treated within one structural unit and the target is left as an
independently recoverable segment of information in a record. Recall that
it is supposed that it is always much more difficult to retrieve an element
of a segment, to be analytic, than it is to retrieve a whole segment.
Certain aspects of metacontrast masking can be viewed in terms of
adding the spatial parsing aspect of event segregation to temporal parsing.
In most cases of pattern masking by structure, target and mask contours
overlap in space; in metacontrast they do not but are juxtaposed (e.g.,
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
265
where a ring masks an adjacent inner disc). Metacontrast is less effective
and requires greater proximity of target and mask contours nearer to the
fovea (Kolers & Rosner, 1960). Masking by overlapping structure can be
seen as a more general and effective case of metacontrast which continues
the relationship between contour separation and masking into the fovea.
That is, the more retinally eccentric a pattern’s projection on the retina,
the greater will be the area surrounding it in which it is susceptible to
interference by confusion of its contour with that of another pattern. In
terms of recovery, the relative resolution power of the fovea requires
maximum spatial (and temporal?) confusability or structural ambiguity,
for parsing into the same segment, and therefore masking, to occur.
Support for this position is given by a study of Lehmkuhle and Fox
(1980). By means of random dot stereograms the relative perceived depth
planes of an inner Landolt C and an outer ring could be manipulated,
where the task was to identify location of the gap in the C. When the mask
appeared farther away than the target, masking declined with increasing
apparent depth separation. Plausibly the “inner” C was more segregated
as an independent spatial event or object.
We must now ask what governs the recovery of events treated as belonging to the same structural unit. We will attempt to apply to masking
Hochberg’s (1970) organizational approach to selective attention, which
supposes that the perceiver’s aim is to make sense of as much of a stream
or image as possible. Identifying what is there involves imposing a structural description. Until this is done all that a segment consists of is a set of
elements. Imposing a structural description on these elements amounts to
constructing a “figure,” and if not all the elements are accountable in the
“figure” this would leave a “ground.” (However, note that in normal
Gestalt terminology both figure and ground are consciously perceived as
such. Here we are referring to figure and ground in the set of elements in
the nonconscious record, only the figure of which can become conscious.)
Two main principles of recovery, plus a combinatorial principle are proposed to account for masking effects. (1) All other things being equal,
within a temporal segment relative proximity in time (recency) or space
will differentiate elements. If a subset of elements are “stronger” due to
recency, it will be relatively privileged in being given Iigural status and
capturing recovery. (Indeed recency might well be thought of as a general
figural principle in the domain of time.) (2) If one set of elements corresponds to a more economical description, or to a higher level description,
or to something more “expected” (i.e., a stronger perceptual hypothesis)
then it will be relatively privileged in being granted figural status, and
would require greater competition (e.g., from recency or proximity) from
other candidate element sets for recovery. (3) Provided that element subsets are not too separated by factors such as time or space, then if ele-
266
ANTHONY
J. MARCEL
ments from two objectively separate event sets can be fused to produce a
figure which itself provides a more economical, expected, or high-level
structural description than either of the two objective element sets, then
that may outweigh physical factors to achieve figural status, leaving the
other elements as ground, and thus unrecovered.
Do these principles apply to observed data on masking? The first notion
is indeed exemplified in that in central masking the later stimulus is
privileged and masks the earlier stimulus (Turvey, 1973). The nearer it is
to the target in time (up to a point), the more they are in competition and
the less chance will the first stimulus have of recovery. In the present
terms, an interesting spatial variant of this is provided by Lehmkuhle and
Fox’s (1980) study, referred to above. In stereoscopic metacontrast
masking where target and mask vary in their perceived depth plane, the
stimulus perceived more proximally was always relatively privileged. The
more perceptually proximal was the inner Landolt C, the more it evaded
masking: i.e., it was more likely to be given figural status. When the outer
ring was more proximal, masking stayed at its severest or became even
more severe with greater relative proximity.
The second organizational factor is illustrated by Jacobson’s work described above and by that of Taylor and Chabot (1978). In Jacobson’s
experiments the higher the level or the greater economy of description
applicable to the second stimulus, the greater was its power to mask the
first stimulus and longer intervals were needed to render them out of
competition. In Taylor and Chabot’s experiment, single letters, five-letter
pseudowords and common five-letter words were followed by a blank
flash, strings of overlapped pairs of letters, pseudowords, or words.
Backward masking was more pronounced when a target was followed by
a stimulus describable at the same level than when it was followed by a
stimulus describable only at a lower level than itself (a word followed by a
word versus a pseudoword), and masking was even more pronounced
when the following stimulus was describable at a higher level than the
preceding stimulus (a pseudoword followed by a word versus a pseudoword). The fact that strings of overlapped letter pairs was the most effective mask can be ascribed to its providing the most economical description as a redundantly repeated pattern.
If two patterns are not sufficiently separated by physical differences in
their structure (e.g., color) or recency then elements from each may be
integrated in the process of recovery to form a new figure. Thus, in
Schultz and Eriksen’s (1977) study (described above) elements of the
enhancement mask were grouped as part of the contours of the target,
permitting a better graphic structural description of the digits. In unpublished experiments by Creighton (Note 1) if a letter string or word is
followed by another letter string or word such that only one word is
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
267
consciously perceived, letters from approximately equivalent positions in
the two were often substituted to form a new word if possible (e.g., given
STEER followed by CHEAP, the percept might be SHEEP). Out of all
the lexically permissible combinations of the letters, what was perceived
tended to be of higher frequency or imageability. Thus various structural
hypotheses may be suggested by the output of processes of perceptual
analysis and the most lexically favored might be sought first in recovery of
data to match it.
Several phenomena which are often thought to be governed by
peripheral mechanisms show equivalent effects to those above. When
different patterns are presented to each eye, instead of complete binocular
rivalry, one may find different elements from each combined into a new
unitary percept (Treisman, 1962). Walker (1978) has already noted the
similarity of binocular interactions to the central as opposed to peripheral
nature of dichoptic masking indicated by Turvey (1973). But we can go
further. First, simultaneous dichoptic presentation, as in both binocular
rivalry procedures and stereoptical depth disparity, directly puts two element sets into one temporal segment and perceptual hypotheses will influence recovery and synthesis toward the most convenient structural
combination of elements. Second, as Walker notes, information from the
suppressed eye affects reports of the presentation to the dominant eye,
even though nothing of the suppressed eye is conscious.
Auditory analogs to these phenomena in dichotic listening have been
discussed by Cutting (1976). In phonetic feature fusion, when /ba/ and /ta/
are presented to different ears, if they are not heard separately nor compete, the most frequent percepts are /da/ and /pa/, where the voicing value
of one stimulus is combined with the placing value of the other. In
phonological fusion, when “back” and “rack” are presented they do not
fuse, but when “back” and “lack” are presented they may be heard as
“black”. Several points are worth making about auditory fusions. First,
physical parameters such as relative onset time and relative intensity
affect the probability of fusion. Second, linguistic rules govern what is
heard in phonological fusion: while “tass” + “tack” can yield either
“task” or “tacks,” “ back” + “lack” yields “black” but not “lback.”
This constraint on consonant cluster order in English also affects the
probability of fusion for different asynchronies when, e.g., /da/ leads /ra/
as opposed to the reverse. Third, when low-level, peripheral features fuse
(waveform and acoustic features) it is by integration (e.g., for sound
localization), but when high-level features fuse it is by disruption and
recombination of linguistic features. This parallels exactly the distinction
between peripheral and central visual masking espoused here.
268
ANTHONY
J. MARCEL
Application to Type B Masking Functions
Interestingly, the current proposal deals with an aspect of masking for
which it was not explicitly designed to account. This is the occurrence of
Type B masking functions. If the magnitude of masking is plotted as a
function of SOA, one of two types of function is generally obtained (Kolers, 1962). The Type A function is a monotonic inverse relationship between SOA and magnitude of masking effect, where maximum masking is
obtained at 0 msec SOA. Type A functions are obtained with both forward
and backward masking, with both noise and pattern masks, and usually
when mask energy is greater than target energy (Turvey, 1973). Thus
Type A functions can be associated with Peripheral masking. The Type B
function is a nonmonotonic (inverted U-shaped) relationship between
SOA and magnitude of masking, where maximum masking is obtained at
some positive value of SOA, i.e., where mask onset lags behind target
onset. Type B functions have not been obtained with noise masks but
have been obtained with both nonoverlapping (metacontrast) masks
(Bridgeman, 1971) and overlapping pattern masks (Purcell & Stewart,
1970; Weisstein, 1971; Turvey, 1973). They are obtainable when target
energy exceeds mask energy (Turvey, 1973) and only with backward
masking. They are found with both monoptic (Weisstein, 1971; Turvey,
1973) and dichoptic presentation (Weisstein, 1971). Plainly Type B functions are associated with Central masking.2
The present account of central masking deals with Type B functions as
follows. As SOA is initially reduced the target and mask elements are
treated with increasing probability as belonging to the same parsed temporal segment of the visual stream. The relative recency of the mask
elements gives them privileged status for recovery. However, as SOA is
further reduced the relative recency of the mask elements decreases,
decreasingly outweighing any privilege given to the target by the target’s
potential level or economy of description. It is extremely important in this
context that in virtually all studies obtaining Type B functions with pattern masks, the targets have been describable at higher levels or more
economically than the masks. That is, the targets have typically consisted
of letters, words, or some recognizable figure, while the masks have
typically consisted of a nonfigural set of elements or a row of repeated
2 However, with monoptic presentation and when target energy exceeds mask energy, a
U-shaped function is partially due to peripheral masking. With small SOAs the target forward masks the following “mask” peripherally; with increasing SOA forward masking
decreases until the mask backward masks the target centrally; this in turn decreases as SOA
is further increased (Turvey, 1973, Experiment XVIII).
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
269
letters. Thus at lower SOAs, level of description of target elements will be
less outweighed by, and will begin itself to outweigh, recency of mask
elements in a competition for recovery.
The apparent exceptions to this are when Type B functions are obtained
in metacontrast, where target and mask have equal tigural properties,
such as a disc and a ring. However, as Breitmeyer and Ganz (1976, pp.
3-4) point out, obtaining Type B functions in metacontrast is dependent
on the measure used. They also point out that when Type B functions are
obtained for form in metacontrast, it is at those SOAs that perception of
motion (from target to mask) is obtained. This fits in well with the current
conception, since if target and mask are adjacent and figurally compatible,
the most economical description of that segment of the visual stream
where they are parsed together and where relative recency begins to
separate and dominate less is in terms of movement.
Accounting for Masking Effects in the Preceding Paper
At this point an interpretation of the masking effects in the preceding
paper (Marcel, 1983) is due. Those in Experiments 3, 4, and 5 present no
great problem. At certain SOAs subjects were unable to detect reliably
the presence of a word preceding the mask, and yet the effects of lexical
representation of that word were reliable. Since no other figural event
closely preceded the word or followed the mask, the present position
would suggest that in the relevant temporal portion of the graphic record,
at the relevant SOAs the word and mask are parsed into the same segment
and the relative recency of the mask is sufficient to grant it tigural status
for recovery. This of course has no effect on the further nonconscious
processing of the results of graphic analysis, which will produce priming
effects.
If the order of effects of SOA reduction in Experiment 1 are valid, an
interpretation of them also needs to be offered. As SOA was reduced,
detection suffered first, followed by graphic similarity judgments, and
finally by semantic similarity judgments. (This of course was only true of
one subgroup of subjects, termed “passive” subjects.) Before addressing
this issue a comment should be made on the basis on which subjects made
their responses, which Experiment 2 went some way to clarifying. In
Experiment 1, since the pairs of choice stimuli varied only on the relevant
dimensions, if residual graphic and semantic activation from the masked
stimulus combined with that from each choice stimulus, it would produce
differential activation on only the relevant dimension. If a stimulus yielding greater activation elicits an orientation response, the most similar
stimulus would elicit such a response and the subject could respond on the
basis of that. In Experiment 2, one member of each pair of choice stimuli
was similar graphically and dissimilar semantically to the masked word
270
ANTHONY
J. MARCEL
and vice versa for the other member of the pair. Reliable judgments on
each of the required dimensions were no longer able to be made even by
“passive” subjects who had done so in Experiment 1. Therefore it appears that the “passive” subjects were successful in Experiment 1 not
because they could be intentionally and selectively sensitive to graphic
and semantic information from the masked word, but because they relied
on a nonconsciously produced orientation response to the more activated
choice word.
An account can now be offered of the SOA effects. As SOA is lowered
the first effect would be an increased probability that the records of target
and mask would be treated in recovery as one event set in time, recency
favoring recovery of the mask elements. This again would impair phenomenal awareness of the premask word but would leave intact nonrecovered results of its visual analysis. A following choice word of greater
graphic similarity to the masked word would benefit primarily from activation of its spatial locational distribution resulting from low spatial frequency analysis (Breitmeyer & Ganz, 1976) of the masked word, and also
from activation of its orthographic composition from high spatial frequency analysis of the masked word. The latter analysis would provide
data for lexical processing whose resultant priming would benefit a word
of greater semantic similarity.
As SOA is further reduced the results of both high and low spatial
frequency analysis (not their records) each becomes less temporally separate for target and mask. This will have quite different effects on information about spatial locational distribution and on information about contour
structure. Locational distribution information will become decreasingly
specific. However, if we posit letter specific graphemic recognition devices (cf. Selfridge, 1959), contour information from the target, even
though accompanied by that from the mask, will still support a graphemic
description of its letters. It would make no sense for such long-term
specifications to exist for locational distributions, and so there would be
no top-down process to pick out the global low-frequency shape distribution of the target from that of the mask. Thus a second stage of SOA
reduction would impair the representational basis of the activation underlying graphic similarity choices, but would leave intact that of lexical
processing and thus of the activation underlying semantic similarity
choices.
Finally, at lower SOAs, since target duration equalled SOA (and
masking was binocular and not dichoptic), the increasingly greater relative mask energy and shorter SOA would lead to impairment of perceptual
analyses themselves by peripheral masking. This would naturally preclude graphic processing and therefore graphemic and lexical categorization and hence any semantic effects.
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
271
Concluding Comments on Masking
Conscious experience, of a “sensible” unitary world, is dependent
upon the ability to match hypotheses of various kinds, produced by the
output of automatic perceptual processors, with the records of certain of
those outputs. Central masking interferes with the recovery of that information by producing competition with it for tigural status in the to-berecovered records. The principles, enumerated above, according to which
it does so follow from the intensional goal of consciousness, event perception, and are classic Gestalt principles (Koffka, 1935). However, one
point of this paper is that Gestalt principles apply to phenomenal experience and not to perceptual processing itself. On the whole central masking
does not affect nonconscious processes of perception nor their effects.
But at severely low SOAs it is suggested that even nonenergy masking can
affect some nonconscious processes, principally those relying on low
spatial frequency analysis, if the masking is not dichoptic.
BREAKDOWNS
IN RECOVERY
AND SYNTHESIS
The current approach postulates different aspects of conscious experience and separate underlying processes. Some validation is required (a) of
the distinction between these aspects of phenomenology and (b) of the
distinction between these processes. In this section various normal and
clinical-neuropsychological
phenomena will be used to illustrate the
phenomenological consequences of changes in and impairments to different aspects of the normally functioning system. For example, in postulating recovery and synthesis as separate processes, there ought to be differential consequences of breakdowns in one and not the other and of the
reverse.
Three phenomenological aspects of normal conscious experience have
been distinguished: (a) awareness or acknowledgment of the existence of
(discrete) entities or events; (b) the unitariness of percepts, where objects
and scenes cohere and where appropriate dimensional values of locations
or segmentations of the visual (or auditory, etc.) field are put together
after independent analysis by specialist subsystems; (c) ability to selectively attend to or recover different types or levels of information about
any one event or segment. The proposed underlying mechanism is that
sensory analyses generate perceptual hypotheses which guide the recovery of information from records and its synthesis, and against one of
which such information is verified. Broadly speaking, in terms of the
proposed mechanisms, consciousness might be subject to limitations, impairments, or failures in two main areas: (a) in recovery of various levels and
types of records (restrictions in accessibility, limitations in capacity, or
total or near-total failure), or (b) in synthesizing recovered information.
272
ANTHONY
J. MARCEL
Each of these problems would have relatively specific consequences related to the three aspects of consciousness.
In what follows we will examine normal, experimental, and clinical
neuropsychological examples of the consequences of each of such limitations or failures, as permitted by the present approach. Several points
should be made here. First, in the clinical cases we are not primarily
concerned with functional neurophysiology, but rather with natural fractionation of intentional behavior or phenomenal experience: that is, with
validation and illustration of the proposed scheme by means of dissociation. Second, what will be discussed are not so much clearcut predictions
as explications of clinical and normal states in terms of the present approach. The clinical conditions mentioned are not meant to be exhaustive,
only illustrative. Some disorders will be mentioned in only one modality
either because of lack of appropriate data or because of space considerations. Third, due to the Identity Assumption many normal and clinical
states are usually interpreted in terms of the presence, absence, or level of
functioning of a structure or process, independently of the role of the
agent, the conscious perceiver-actor. Many of these will here be taken to
reflect aspects of consciousness and the agent’s ability to access the respective structures or processes in different ways. In fact some of these
phenomena can be seen as functional parallels to masking effects. However, this is not meant to imply that all or even most neuropsychological
disorders are primarily disorders of consciousness. Many conditions do
reflect basic functioning, and were it not so the present position would be
meaningless. Fourth, the interpretation of clinical syndromes here is
prototypical;
due to the nature of neurological impairment and
neuroanatomy, “pure” syndromes are rare. Fifth, it is assumed that
modular impairments will often not be reflected directly in behavior, but
will be masked by individuals’ attempts to cope with impairment in order
to fulfill overall pragmatic or social goals.
(a) Recovery
(i) Impaired recovery of lower levels. Selective access to one aspect or
level of description is what underlies the ability to choose that aspect or
level to control categories of behavior, i.e., what the subject is instructed
to do in the vast majority of perceptual experiments. It is normally the
case that we operate consciously at the highest or most functionally useful
level-in language at the level of communicative pragmatics, in vision at
the level of objects, egospace, and action. Donaldson (1978) has pointed
out that the academic aim of Western schooling is the ability to detach
one’s consciousness from overall goals and determination by the outer
world and to attend analytically. As she has discussed at some length,
many of the difficulties encountered by children at school consist of learn-
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
273
ing to address themselves analytically to specific aspects or low levels of
representation of their own percepts and actions. One example of this (see
Mattingly, 1980) is that people who perceive and produce speech adequately may nonetheless have difficulty in attending to the phonetic segmentation of speech (locutionary awareness). Though we acquire the
ability to address ourselves analytically to low-level representations of
our percepts, we may lose these abilities with neurological trauma. Thus
Luria (1947, 1950) describes patients who are quite capable of normal
speech perception and production but are unable to break a word down
into its succession of sounds. Goldstein (1948) emphasizes in aphasia the
breakdown of the “abstract attitude” (almost exactly the converse of
what Donaldson describes as being acquired at school) at various levels.
Thus consideration and production of isolated sounds is frequently more
disturbed than their utilization in perception and production of words, and
the same is so for isolated words compared to their perception or production in sentences. In vision, apperceptive object agnosics have difficulty
in abstracting visual qualities from the object or cannot recognize an
object out of context while they often have little difficulty dealing appropriately with the same object in context. In the sphere of action,
ideomotor apraxia (Brain, 1961; Hecaen, 1972) provides examples of
patients who, with no comprehension problem, can spontaneously perform complex acts but cannot produce their components to order. For
example a patient who cannot comply with the request to “hammer the
nail into the wall,” does so when asked at home to hang up a picture. In all
these cases, just as in Experiment 1 of the preceding paper (Marcel, 1983),
intentional access to processes and representations may break down for
the nominally more simple before the more complex. But this classification from simple to complex is only appropriate in an atomistic sense.
From the phenomenological point of view it is the reverse. The description
of the visual array at the scene level is in fact the most accessible, as is the
description of movements at the level of an action.
(ii) Impaired recovery of higher levels. Developmentally, increasingly
abstract or higher level ways of structuring experience will become
operative as the corresponding constructs are acquired. Clinically, the
unavailability of normally high-level hypotheses will lead to a lack of the
experience of the meaning or value of percepts, but will leave intact the
nonconscious processing of such meaning. The patient will be reduced to
utilizing hypotheses at lower levels, and therefore unity per se will not be
lost. (Note that we are discussing the irrecoverability of higher level descriptions. It is plausible that in the dementias these levels of representation themselves disintegrate.)
Classically this description fits the associative agnosias (Lissauer, 1890).
The patient appears to have an adequate phenomenal apprehension of the
274
ANTHONY
J. MARCEL
full physical or sensory structure and nature of an object or sound, but
still does not appreciate “what” it is. This can occur in one modality at a
time. The most telling experimental investigations of associative object
agnosia so far are those by Warrington (Taylor & Warrington, 1971; Warrington, 1975). That this is a problem of consciousness rather than
analysis itself is suggested by data in the latter paper where categories that
could not be addressed consciously nevertheless appear to have affected
the patients’ behavior. Words that could not be understood nor distinguished from pseudowords produced better immediate memory performance than the pseudowords. Unfortunately since the Identity Assumption has been tacitly and strongly held, adequate data of this type has yet
to be gathered.
However, there is one example where relevant data have been obtained. Aphasic patients who cannot comprehend the meaning of many
words can often distinguish these words from pseudowords and read or
repeat them. Thus these people appear to consciously apprehend the
sensory and lexical characteristics of the words but not their semantics.
Nevertheless, semantic priming effects in both visual and auditory lexical
decision are obtained from those same words (Milberg & Blumstein, 1981;
Blumstein, Milberg, & Shrier, in press). Thus, in Milberg and Blumstein’s
words, “it would seem that the failure of patients to perform tasks requiring metalinguistic judgments or conscious semantic decisions reflects a deficit in accessing and operating on semantic properties of the
lexicon and not an impairment in the underlying organization of the
semantic system used.” (1981, p. 381)
It is pertinent to the present analysis in terms of recovery that Goldstein
(1948) and Hecaen (1972) note that in such cases an object’s identity or
significance may be lost only to intentional experience, e.g., deliberate
attempts to comment on it or respond on its basis during testing. Its significance may still be operative in the normal course of events when the
patient is not attending to it.
It is difficult to find corresponding phenomena in normal people precisely because our consciousness is normally directed at higher levels of
representation. However, when the processes which yield higher levels
of representation are not fully automated we may see the equivalent.
For example the less skilled reader in devoting more attention to coding
processes may lose consciousness of the meaning. Another example
may be semantic satiation, where repetition of a word leaves intact
consciousness of its physical aspects, which may even lose their coherence, but prevents apprehension of its meaning or lexical unity. It is
arguable, though highly speculative, that in studies of masking such as
those by Jacobson (1974, 1978) the additional interfering effects of a mask
having the same lexical status as the target derive from interference with
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
275
recovery of records of lexical analysis of the target over and above those
of specific graphic features.
(iii) Capacity limitations on recovery or ver$cation.
A plausible reflection of normal capacity limitations is the initially rapid rate of gain of
information from briefly exposed displays dependent on the SOA before a
mask (Sperling, 1963; Merikle, Coltheart, & Lowe, 1971). Clinical manifestation of abnormal limitation is to be found in one variety of simultanagnosia, where the patient seems to have highly spatially restricted
attention. In a scene, objects and regions can clearly be parsed and synthesized but only one achieves consciousness at a time. As a typical
example, Wolpert (1924) reported a patient who when confronted with a
picture of a dog crossing a street described the scene as “It’s a dog.”
Kinsbourne and Warrington (1962, 1963) have shown that such patients
have a particularly limited rate of gain of information from tachistoscopic
displays. Another clinical manifestation of a limit on recovery, more severe, is plausibly provided by cases of sensory extinction or inattention
(Bender, 1952). When a single visual stimulus is presented in either visual
hemifield the patient demonstrates awareness of it. However, if the two
are bilaterally displayed simultaneously the patient only demonstrates
awareness of that on one (consistent) side, denying the other’s presence.
The same phenomenon exists for audition, touch, and other sense modalities, though whether it concerns side of body or of space is unclear.
Strikingly, if patients are forced to make same-different judgments, e.g.,
with words or pictures, much as they find the task nonsensical, they
perform at 88-100% accuracy (Volpe, Ledoux, & Gazzaniga, 1979).
Clearly the nonconscious stimulus has been processed, yet it seems as if a
permanent (lateral) bias in recovery is operating, where a particular description can only be used to recover one exemplar in the record at a time
and, if there is competition, one side is privileged.
(iv) General failure ofrecovery.
The most serious and dramatic impairment to recovery will be failure to recover at all. If we cannot recover any
of the records pertaining to physical aspects of a discrete event then we
would have no basis for acknowledging its presence. If we cannot recover
records of physical qualities at all, the approach outlined here suggests
that we would have no awareness within the relevant modality. Patients
with damage to the geniculostriate visual pathways exhibit cortical blindness (most often restricted to one hemisphere and therefore the blindness
is hemianopic, restricted to one hemifield).
Such patients are
phenomenologically blind in the scotomatous area, at least for static visual arrays, and cannot detect or comment on visual stimuli, yet they can
localize them via eye movements (Poppel, Held & Frost, 1973) and
276
ANTHONY
J. MARCEL
pointing (Weiskrantz et al., 1974), and, while they claim to be guessing,
they can grasp objects correctly and make successful forced choices in at
least gross shape discriminations (Weiskrantz et al., 1974). For present
purposes, it is important to draw a contrast in the interpretations of this
syndrome, now known as Blindsight. It has usually been interpreted in
terms of a dissociation of visual function. The division drawn by
Schneider (1967), Trevarthen (1968), and Humphrey (1974) is between the
geniculo-occipital
system (damaged in such patients) and the pulvinar-collicular
(tectal) system whose cortical projections are parietal
(intact in such patients). Authors have tended to associate the former with
treatment of sensory qualities of parts of the optic array, enabling identification, and the latter with treatment of motion and location in terms of
some map of the visual world. Thus the syndrome has been seen as a
dissociation of function. However, this does not deal with the primary
characteristic, which is in terms of a lack of phenomenal visual experience. In contrast one can view the problem as one of visual consciousness. In this context it is relevant to mention some current work on two
such hemianopic patients by the present author and Dr. A. Wilkins (Marcel, 1982, Note 2). In addition to replicating earlier work, we have found
(a) better than chance preparatory adjustments of the wrist, fingers, and
arm in reaching for objects of differing shape, orientation, size, location,
and distance in the blind field; (b) the ability to produce phenomenal
percepts in the blind field by afterimages, if accompanied by figurally and
locationally related afterimages in the sighted field; (c) most dramatically,
the ability to bias the interpretation of an auditorily presented polysemous
word (BANK) by a preceding 20-msec presentation of a word related to
one of its meanings (RIVER/MONEY) in the blind field. These data imply
that sensory qualities have indeed been analyzed but the problem is one of
recovery and thus of awareness. The neuropsychological implication of
this is that occipital cortex may deal not so much with analysis of sensory
qualities comprising form and identity but with their recovery for awareness. Weiskrantz (1977) has tentatively extended this argument to the
cortex as a whole, with special reference to memory as well as vision.
With regard to memory, Squire has recently (1982) reviewed the many
findings that amnesic patients demonstrate learning or retention behaviorally without being able to reflect it in their verbal reports or recognize their memories as memories. One view of this is that amnesic patients do not have (conscious or intentional) access to their memories. In
terms of the present approach we can suggest that while past experience
automatically and nonconsciously has its effects, the amnesic person is
unable to recover the records of such experience from temporal episodes
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
277
other than the current one. Thus the person would have no phenomenal
experience of thoughts as memories, i.e., as being associated with a particular external source or temporal episode.
(b) Synthesis
It has been suggested here that synthesis is of two types. Synthesis of a
particular level of description of an object with its constituent features is
achieved via the specification of those features and their structural relations in the perceptual hypothesis pertaining to the object. Synthesis of
the particular values of those features with the features is achieved via a
spatial map. Selective impairment of or interference with the ability to
synthesize unitary percepts could arise in two ways and should manifest
itself in different ways. If the record of the spatial source of independently
analyzed aspects of the environment is inaccessible, then individual feature values would not be able to be reallocated veridically. If higher level
perceptual hypotheses are inaccessible then their constituent features
could not be put together appropriately. In both cases individual features
and their values would be recoverable and the perceiver would be aware
of them but their integration would be impaired. Note that we are here
discussing integration only relative to a single structuring opted for in the
achievement of a conscious percept, and not the multiple integrations
carried out nonconsciously which suggest the perceptual hypotheses in
the first place. We take issue with Treisman and Gelade’s (1980) view that
conscious attention is the only process responsible for integration. If it
were, categorical percepts that depend upon feature conjunctions would
not be possible nonconsciously. As the studies in the preceding paper
(Marcel, 1983) and those of McCauley, Parmelee, Sperber, & Carr (1980)
indicate, strokes and letters are integrated to form words and picture
elements are integrated to form scenes at a nonconscious level.
There are two clinical syndromes which appear to correspond to the
two ways in which synthesis can break down.3 Balint’s syndrome
(Balint, 1909; for reviews see Hecaen & Albert, 1978; Rubens, 1979) is
usually characterized by the co-occurrence of several symptoms. The
patient is unable to reorient their gaze at will and is also impaired in
localizing visually perceived objects. These symptoms have been related
3 There are particularly
diffkult
problems in dealing with this area of clinical neuropsychology. First, authors disagree as to whether patients fall into one category or another.
This disagreement is often due to severity of impairments and whether the impairments are
unilateral or bilateral. Second, which tests are carried out differs from one study to another.
So it is difftcult to decide which impairments co-occur with each other and which dissociate
from others. Further, careful records of the phenomenal experience of patients, which is
central here, are few.
278
ANTHONY
J. MARCEL
to a deficit in the visual representation of egocentric spatial relations. The
patient also has a severe inability to perceive more than one object at a
time. Luria (1959) reported a case where if a star made of two overlapping
triangles was presented it was seen as a star, but when the triangles were
of different colors, only one triangle was perceived. Further, something
not usually examined, aspects of an object or feature dissociate phenomenally from the object, are considered one at a time, and if two objects
are present may be wrongly allocated. When a woman examined by Tyler
(1968) was shown a dollar bill and a cup, she saw “a cup with a picture of
Washington on it.” Adler (1944) showed a patient two pictures of a boy
with a toy boat. She denied that the boat in the two pictures was the same
boat because of its different color. But when referring to the color of the
boat she was actually pointing to the blue water. Rafal (personal communication) reports a patient who when shown a yellow water pitcher said
that it contained lemonade, apparently misattributing the color to the
liquid. The same patient, when shown colored shapes, could correctly
report one dimension at a time, but when asked to report both shape and
color made more than 50% errors, where one dimension was correct and
the other not. It is of interest here that in such cases dissolution of the
arbitrary values of objects or features co-occurs with a disturbance of the
representation of space and that patients can only see one object at a time,
needing a long period between successive presentations of single objects.
The clinical condition which most closely approximates the dissolution
of features or parts of objects is what Liepmann (1908) called disjunctive
agnosia, characterized as fragmentation of representations into their elements. The phenomenal experience of visual fragmentation in a severe
case is well exemplified by a patient who premorbidly had been a
psychologist skilled in introspection. Extracts from the transcript of an
interview with him appear in the Appendix. These extracts make it obvious that he was simultaneously aware of, able to recover, all features
present in a static scene together with their correct arbitrary values, but
could not put them spatially together as objects. It is worth contrasting
this with Balint’s syndrome. In disjunctive agnosia the patient is simultaneously aware of parts of several objects and their correct values (e.g.,
color) but may not see the parts correctly conjoined, i.e., bits of objects
are not seen at the level of objects. In Balint’s syndrome the patient can
see objects correctly as objects but only one at a time, and the arbitrary
values of each object may be incorrectly conjoined to the object. It is
interesting that in the case of disjunctive agnosia, synthesis of parts of
objects appears from the transcript to be aided by movement in the visual
field. Psychologists of widely differing orientations (Koffka, 1935;
Johansson, 1974; Turvey, 1977b) have argued that movement provides a
more primary cue to segmentation of the visual world into objects than
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
279
static information. Thus, if the top-down synthesis derived from the
structural description constituting a perceptual hypothesis is not available, relative motion and parallax would serve to synthesize the elements
recovered from records. Indeed, if structural descriptions of objects are
not available to phenomenally segment a static scene, we might well
expect other Gestalt principles of organization (e.g., common color) to
phenomenally unitize the elements and that certain principles (common
movement) would override others.
The functional equivalent of each of these two kinds of breakdown in
synthesis can be found in studies on normal people. Synthesis of arbitrary, episodic feature values has been investigated in Treisman’s recent
work (Treisman, Sykes, and Gelade, 1977; Treisman and Gelade, 1980),
using a variety of procedures. The studies suggest that allocation of values
to spatially defined sources is crucial, that this process is capacity-limited
and takes time, requiring serial processing of each object-defined spatial
location for conjunction of the feature values, and that segments of the
visual field which have not yet been attended are subject to false
phenomenological conjunctions. We can thus see that the inability to
consider separate feature dimensions simultaneously and the false conjunctions of these values found in Balint’s syndrome are plausibly an
exaggeration of the normal state of affairs. In very rapid serial visual
presentation to normal subjects, not only are some items not recovered
but the particular values of arbitrary dimensions are often synthesized
with the wrong object. In numerous studies aimed at short-term visual
memory, arbitrary displays are presented briefly and viewers are asked to
report selectively the objects according to color, size, etc. This requires
the synthesis of object identities with their episodic features.
Experimental realizations of failure to accurately synthesize constituent
features are plausibly to be found in studies by Allport (1977) and Shallice
and McGill (1978). In both of these studies several words were presented
simultaneously followed by a pattern mask. The graphical qualities and
relations between target and mask were comparable with those in my own
studies (Marcel, 1983)except that the SOAs were relatively longer. In attempting to report the words, subjects tended to recombine letters and letter groups
from separate words to form a different word while maintaining the correct structural location from each of the source words. Examples from
Allport include the stimulus words rust and vent leading to responses
“runt” and “vest,” and tab and cud leading to “cab,” “cub,” or “tub.”
While lexical or orthographic constraints may have operated in the generation of perceptual hypotheses to determine subjects’ percepts or responses, one of the effects of masking was nonetheless to selectively
disturb that which would allow veridical recombination of correctly identified segments. As Allport (1977) has hinted, models of visual recognition
which lack such a mechanism are seriously deficient. That combination of
280
ANTHONY
J. MARCEL
the elements was veridically achieved nonconsciously is suggested by the
fact that Allport found semantic effects from the presented words even
though they were incorrectly synthesized consciously.
The difference between these studies and disjunctive agnosia is as follows. Visual structural descriptions specified by object-level perceptual
hypotheses (words) are available to the normal person but cannot be
verified. Therefore the spatial segments will get combined at that level,
though often wrongly. In visual disjunctive agnosics, the structural descriptions specified by object-level perceptual hypotheses are not accessible in the visual domain. Therefore in static displays the segmentable
features will not get combined at that level, and will be subject to lower
level principles (color, movement) for their organization.
REINTERPRETATION
OF PERCEPTUAL
PHENOMENA
In the introduction to the preceding experimental paper (Marcel, 1983)
an indication was given of how the Assumption of Identity (between
processing and the conscious representations upon which responses are
based) has guided interpretations of perceptual phenomena in terms of
synthetic or analytic structurings of perceptual microgenesis. In falsifying
that assumption, two general principles have been adduced which will
affect any reinterpretation of particular kinds of phenomena, and which
should be born in mind in any attempt to experimentally address conscious or intentional processes. The first is that perceptual experience and
reports and intentional responses are based on recovered information and
reflect characteristics of recovery. The second is that conscious perception is a “late” stage: nonconscious processes and representations precede and affect conscious representation and characteristics of intentional
responses.
In this section, three new paradigm assumptions will be illustrated by
reference to their implications for reinterpretation
of perceptual
phenomena: the nonidentity of processing and experience, the concept of
recovery, and the effects of automatic preconscious processes. These
three concepts correspond to different aspects of Figs. 3 and 4: (a) the
functional separation of the records from the results of processing; (b) the
role of perceptual hypotheses and the directional property of recoveryrecovery works top-down, the records of the more analytic representations can be addressed via records of those derived from them but
not vice versa; (c) the automatic activational properties of unconsciously
achieved descriptions.
I. Nonidentity
Phenomenal
of Process and Experience-Iconic
Persistence
Memory
and
The historical core of this and the preceding paper is the phenomenon
of visible persistence and the construct of short-term visual memory. In
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
281
the introduction to the preceding paper it was mentioned that what masking has been assumed to affect is a representational stage termed iconic
memory. This concept has assumed an identity between neural persistence, phenomenal visible persistence, and informational persistence
after stimulus offset. Plainly the work reported in the preceding paper
calls the concept into question and the ideas proposed in this paper offer
an alternative approach to the relevant phenomena. Recently Coltheart
(1980) has provided a searching review of this area and his conclusions are
directly in line with the views here. Thus it seems most appropriate to
summarize briefly his arguments; they provide a good example of the
direction urged by this paper.
As noted above, the three forms of evidence of visual persistence
(neural, phenomenal, and informational availability) have been assumed to
reflect a single variable: a decaying visual trace that “(1) consists of
afteractivity in the visual system, (2) is visible, and (3) is the source of
visual information in experiments on decaying visual memory.” Coltheart
points out two fundamental properties of visible persistence which are
shown in several different techniques which rely on phenomenal measures: the inverse duration effect (the longer a stimulant lasts, the shorter
is its poststimulus persistence) and the inverse intensity effect (the more
intense the stimulus, the briefer its persistence). He then examines neural
persistence and proposes that at present it is plausible to claim that visible
persistence is produced by neural persistence. However, he is guarded
about this, arguing that the appropriate kind of studies correlating
psychophysics and neurophysiology are sparse and insufficient. His main
point, though, concerns informational persistence. This is critically defined by “partial-report superiority” in the method introduced by Averbath and Coriell(l961) and Sperling (1960), which led to the term “iconic
memory.” Coltheart points out that several studies demonstrate that the
duration of iconic memory or informational persistence, thus defined, is
not inversely related to stimulus duration or stimulus intensity. Thus informational persistence cannot be identified with phenomenal visual persistence. So far, then, this amounts to a further important demonstration
that phenomenology cannot be identified with and bears no simple relation to information processing.
Coltheart makes two further proposals of interest. First, as has been
reiterated throughout the present paper, he derives the implication that
one cannot investigate iconic memory by tasks that require phenomenological judgments. Such “direct methods” tell us instead about visible
persistence. Second, he speculates on an alternative conception of iconic
memory to the accepted one. He proposes that setting up an iconic
memory “consists of temporarily attaching various forms of physical information to a permanently existing entry in the internal lexicon” (or
282
ANTHONY
J. MARCEL
presumably to permanent entries in other descriptive domains). While
there are differences in detail between Coltheart’s and my own conception, this view corresponds almost exactly to the present proposal,
outlined in the section on Recovery, of the process of hypotheses, lexical
or otherwise, being fitted to information of a more analytic, physical, and
episodic nature recovered from the records. In the section on failures of
synthesis it was pointed out that in tasks requiring partial report from brief
presentations the perceiver’s task is to synthesize high-level perceptual
hypotheses with low-level episodic feature values recovered from records. The usual arbitrariness of the features chosen (color, location, size)
is a problem because they cannot be recovered via the high-level records
of the identities. In addition, only one structure can be used at a time for
recovery and synthesis with its constituent features anyway. Thus the
limits of performance in partial report do not reflect what is represented in
“iconic memory” and the duration of that representation, but rather reflect the limits of recovery and synthesis and the duration of various
records.
Note that in both Coltheart’s and my own version, neither (a) the
hypotheses or the records themselves, nor(b) “iconic memory” itself are
conscious. In indicating the application of his conception to perceptual
phenomena, Coltheart gives explanatory force to the notion of selection
from iconic memory as a “late,” postlexical stage. However, he leaves
this largely undeveloped. The purpose of the two subsections following this is
to undertake and illustrate such a development.
Before proceeding, however, one classic phenomenon in the iconic
memory literature seems perhaps an apparent paradox in terms of the
views advanced here. Since ionic memory, as reflected by partial-report
superiority, has been thought of as a sensory precategorical representation, selection from it on the basis of “sensory” features has been supposed possible, but not selection on the basis of postcategorical features
(i.e., those derived from identification). Indeed, partial-report superiority
is found for such characteristics as color and location, but not for alphanumeric class, e.g., reporting letters versus digits (Sperling, 1960; von
Wright, 1968, 1972). Since the present view conceives of partial report as
a postcategorical process, surely it ought to be possible to select on the
basis of identity-derived category? This apparent paradox turns out not
only to be superficial, but also to emphasize one of the central features of
the present view. First, the fact that categorical representations have in
fact been achieved by the stage usually taken to reflect iconic memory, is
witnessed by the experiments reported in the preceding paper (Marcel,
1983) and that by Fowler et al. (1981). The main point, though, is that
practically all the relevant experiments require report as a measure. Report relies on phenomenal percepts and it has been stressed here that
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
283
phenomenal percepts require the recovery of records of episodic or
physical characteristics, and in the case of report of identity what is
required is the synthesis of such information with conceptual representations. Thus even if postcategorical representations have been
achieved, we cannot expect them to be reflected in partial report if to do
so exceeds the temporo-spatial capacities of synthesis and the duration
of the relevant low-level records.
2. Recovery Effects
Let us here attempt to make some use of the notion of recovery of
information. To start with, where different classes of stimuli have appeared to be differentially easy to perceive due to a processing variable,
they may in fact be differentially easy to recover. Consider such factors as
the characteristics of different words (e.g., frequency, imageability), their
context or their visual quality which influence either their tachistoscopic
perception or processing time in, say, lexical decision. Usually effects of
these variables are attributed to the input stage of lexical access. In terms
of the logogen model (Morton, 1968) the effect of such variables is taken
to reflect a logogen’s first threshold and the amount of information needed
to exceed it. According to the present notions, a word may already have
been identified lexically, but its ability to achieve consciousness may be
affected either by its tacit or nonconscious representation in providing
hypotheses for recovery or by factors affecting the recovery and veritication process. Effects of word frequency may be due not to initial access
(e.g., in terms of the threshold) but to the activation properties of that
lexical entry after access. Prior context in the form of lexical/semantic
association could also be affecting recovery of rather than access to a
lexical entry, by influencing the choice and mobilization of candidates for
recovery and verification. Since it is assumed that it is initial access to
the lexicon that is affected by visual degradation by noise, the fact that
degradation interacts with contextual priming in lexical decision has been
taken by Meyer, Schvaneveldt, and Ruddy (1975) to indicate that context
too primarily affects lexical access. Suppose that degradation indeed produces a greater problem for lexical categorization since the output of
graphical analysis will be suboptimal descriptions of letters, but that by
dint of this the record of graphical analysis is also a suboptimal representation. If contextual activation of a lexical entry provides a lexical
hypothesis or heuristic for recovery processes to use in attempting to
match the graphic record, a less well-formed record will suffice.
A notion amenable to this kind of interpretation is that of Holistic perception. The general idea has been that certain phenomena reflect identification of an event in a way other than from its parts and their structure
or order. I know of few models of perceptual processing which could be
284
ANTHONY
J. MARCEL
instantiated computationally that would realize this notion. Two characteristics of recovery have been mentioned, however, which would produce the effects which usually motivate the notion. Firstly, it has been
argued that conscious percepts usually follow the achievement of and
represent the “highest” level of analysis possible. Lower (or component)
levels of analysis are either less recoverable or irrecoverable. Secondly,
given the constructive nature of conscious percepts, lower levels are often
only recoverable through the higher level. We will discuss this below with
regard to specific phenomena. But these principles make it unnecessary to
postulate holistic perception independent of component feature analyses.
As stated earlier, it may well be that representation of a higher level is
only achieved after that of a lower level by being derived from it; but since
this happens before any recovery operation, recovery of lower level information is necessarily influenced by higher levels of representation.
This is one way in which unconscious knowledge affects conscious
knowledge or perception of an event. Two specific examples of phenomena which have invited analytic hierarchical accounts of encoding
are studies of phoneme monitoring and effects of wordness on letter recognition. Specific target phonemes are easier and faster to detect in higher
order speech segments (e.g., words vs. nonwords) and more so when
those segments are more frequent or predictable at the lexical, semantic,
or syntactic levels (see Cutler & Norris, 1978, for review). Does this mean
that the word is in some sense represented prior to the phoneme? Not in
terms of the temporal order of automatic processing, but only in the sense
that lexical and semantic codes are what are normally relevant for conscious representation. For this reason phonemes are recovered through or
with the aid of lexical representation. The more activated that lexical
representation is, for example by contextual priming, the more it will aid
component phoneme recovery or synthesis. Indeed with regard to auditory language perception, Foss and Swinney (1973) have resolved the
problem in the proposed manner by arguing that while linguistic units may
be automatically identified in the order of low (component) to high (combination) levels, the order in which they can be brought into consciousness, and hence responded to, may be to an extent in reverse.
Exactly the same argument applies to Wordness effects in visual letter
discrimination. In Reicher’s (1969) and Wheeler’s (1970) original experiments a letter was presented in the same spatial position either as part of a
word, part of a nonword composed of the same letters, or singly. It was
backward pattern masked and then two probe letters were presented,
above and below the critical spatial position, which could both have made
a word of equal frequency in that position (e.g., if WORD had been
presented then D and K could have been the choice pair). Choices were
more accurate when the letter had been part of a word than otherwise.
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
285
This does not mean that words are easier to see than nonwords (though
that may be true). The appropriate point is that information about the
original letters has to be recovered. At the automatic preconscious stage
letters of both words and nonwords will be represented equally at the
letter or graphemic stage. But letters composing a lexical address will also
produce a lexical representation. That lexical representation will aid recovery of letter information both by providing a route for recovery and an
economic code for their synthesis. Thus recovery is aided in finding the
original letter by its lexical frame. Additionally, or alternatively, the lexical representation will automatically produce greater activation of the
graphemic entries, which will provide a basis for discrimination of the
correct probe. This of course is exactly parallel to the discrimination tasks
of Experiment 1 in the preceding paper. This interpretation is supported
by the fact that Johnston and McClelland (1973) found that the phenomenon is only obtained with backward pattern masking of the original presentation. Energy masking eradicates the effect.
The arguments presented above can also account for similar nonlinguistic visual effects. A line segment is often identified better when presented as part of a drawing that looks unitary and three-dimensional (an
object) than when it is part of a less coherent flat-looking design (Weisstein & Harris, 1974; Womersley, 1977). This parallels the auditory
superiority of phoneme detection in words compared to nonwords and letter-identification superiority in words. Further, Williams and Weisstein
(1978) have shown that a line is better identified when part of a unitary,
apparently three-dimensional drawing than when presented alone. This
parallels the “word-letter”
effect discussed above. Once again, as opposed to postulating an analytic perceptual process of microgenesis, we
can suppose that feature analysis precedes representation of their combination but that more economic descriptions of feature combinations (e.g.,
structural descriptions or descriptions in the object domain) aid in the
recovery of individual component features.
Another way in which preconsciously achieved representations may
help recovery of information has to do with capacity bounds on recovery.
It was proposed that not only does predictability facilitate recovery but
that the more economical the descriptions with which information can be
matched, the more can be transferred to consciousness (as manifested in
working memory paradigms). This applies fairly well to those results
which seem to imply the interaction of higher order effects with iconic
memory. For example, Mewhort (1967) briefly presented pairs of eightletter pseudowords of either zero-order or fourth-order approximation to
English, followed by a cue as to which row of letters to report. He found
that the number of letters correctly reported was increased by the approximation to English, not only of the row reported, but also of the row
286
ANTHONY
J. MARCEL
not reported. According to the classic paradigm assumptions there is a
problem in how what is supposed to be a precategorical store can be
affected by a property of a code derived from it by later processing. It is
even less plausible that such effects should accrue from a part of the
precategorical representation which has not yet received any further processing. However, the account offered above of automatic processing and
aids to recovery avoids such problems. If a structural description can be
given to the to-be-ignored part of the display, it will help in the tigure-ground segmentation necessary to selective recovery and if the subject does start by recovering the “wrong” row, such recovery will be
effected quicker, allowing more time for recovery of the correct row.
Most importantly, conceiving of iconic memory as postcategorical eliminates the paradox and invites the above kind of approach.
Biederman (1972) briefly exposed a real world scene or a jumbled version of the scene. An object was more accurately identified when part of
the former than part of the latter, even when the viewer had previously
been told where to look and for what. It is not clear whether this is best
conceived of as a more macrocosmic exemplar of superiority effects or
whether the more economical description available for the rest of the
scene is responsible. But, whichever is the case, the current view of
recovery makes the phenomenon less paradoxical than views of microgenesis based on the Identity Assumption.
It is worth indicating just how far-reaching the implications of the notion of recovery may be for reinterpretation of perceptual mechanisms.
Take the extreme case of visual acuity. On conventional assumptions all
the data we have lead us to believe that visual resolution at the retinal
level decreases with retinal eccentricity. Conventional interpretation of
physiological
data (e.g., density of innervation)
is validated by
psychophysical data (e.g., spatial frequency resolution, contrast sensitivity functions). Yet psychophysical data for the most part relies on intentional responses based on phenomenal impressions. Consider two anomalous phenomena. First, the meaning of unreportable words presented in
the periphery apparently affects foveally presented tasks (Bradshaw,
1974; Underwood, 1980). Second, in Yarbus’ (1968) studies of picture
scanning, an area of high interest (a human figure in a forest) but of relatively low physical contrast with the surround (in brightness, color, contour) was able to attract an eye movement from far away when areas near
to it had not yet been scanned. Both of these phenomena suggest that the
peripheral stimuli in question had received detailed and high-level
analysis (respectively, lexical categorization and object recognition). Is it
possible that our poor phenomenal perception in the periphery is a problem of recovery rather than of sensory analysis? Phenomenal clarity may
be viewed as a property of figure -ground aspects of attention (see discus-
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
287
sion of masking above), fovea1 projections being privileged. While there
are good ecological reasons for attention to be normally correlated with
direction of gaze, Grindley and Townsend (1968) and Posner (1980) have
shown that the two can be decoupled, peripheral stimuli receiving attention and the perceptual advantages it bestows. Bouma (1970) has shown
that the decrease in recognition with eccentricity interacts with the presence of stimuli flanking the target. Additionally, the separation between
target and flanking stimuli which produces an equivalent degree of lateral
interference appears to be a constant proportion of eccentricity. It is
tempting to speculate on the similarity between this phenomenon and the
relation (indicated in the discussion of masking) between metacontrast
and central masking as a function of eccentricity. That is, visual eccentricity may affect phenomenal experience, via some aspect of recovery
(graphical status in the record?), rather than affecting sensory analysis,
and surrounding nonhomogeneities may have a similar effect to masking.
This speculation may well turn out to be untenable. However, its purpose
is to illustrate the status of conventional assumptions in interpreting perceptual data.
3. Effects of Preconscious
Intentional Processes
Identification
and Activation
on
The third principle which was alluded to at the beginning of this section
is that properties of preconscious representations affect the later processes applied to conscious representations. One instantiation of this is
how “logically” determined responses which are supposed to be based on
a particular categorization can be affected by activation resulting from a
supposedly later stage or totally irrelevant kind of coding. Remember that
we have supposed that preconsciously every possible kind of descriptive
analysis is carried out automatically on a stimulus and to the highest
levels.
In many experiments, characteristics of the response, especially latency, is taken to reflect directly the time taken by perceptual processes
or by logical steps in computing the judgment. However, in many cases
latency may reflect the effect of automatic unconscious processes on
response retrieval or execution, rather than reflect any perceptual process. Consider the negative proximity effect in semantic memory classification (Collins & Loftus, 1975). It is is harder to say “No” to “Is a whale
a fish?” than to “Is a cow a fish?“, supposedly because “whale” is stored
in memory nearer to members of the set of fish, or because it shares more
features with members of that set. Let us suppose that when a memory
location for a stimulus is accessed it primes those other locations with
which it is associated, decreasingly with associative distance or with less
shared features. Those entries prime the response associated with them.
288
ANTHONY
J. MARCEL
Thus probe items near to a set boundary will prime items (and their
responses) in an adjacent set more than would items far from the set
boundary. This will lead to differential response interference. But note
that this account does not assume that latency differences reflect time to
identify the probe item or the logical processes of verification of the
experimental question. It views the source of the effects as quite independent of such processes; it presupposes identification and places the effect
at the stage of response.
This account can also be applied to speeded binary classification of
multidimensional stimuli, and “same-different”
judgments. Reaction
times are often taken to reflect the nature of perceptual judgments. Problems have been encountered since no one comparison model appears to fit
(a) both positive and negative RTs, or(b) the patterns of RTs in classification with both conjunctive and disjunctive criteria (Nickerson, 1972, 1978,
for review). With conjunctive criteria (Red AND circle AND large) which
are equivalent to same-different judgments, “different” or negative RTs
decrease with the number of differing features-consistent with serial
self-terminating feature testing; but positive or “same” RTs are often
considerably shorter than the longest negative RT-which is inconsistent
with the above notion. This has sometimes led to the view that while
negative responses are produced as described, positive responses are the
result of some form of holistic identification. An alternative account views
latency differences as a result of the interaction of the logical decision and
effects resulting from automatic access to a high-level description of the
stimulus. Suppose that, whatever the subject attempts to attend to in
accordance with the task, each stimulus is automatically fully described
and that this description accesses a location in a multidimensional space
representing the stimulus set. Whatever the logical decision about the
present stimulus, response time would inevitably be affected by the degree of priming of the competing responses to the extent of the proximity
of the probes’ representations to those of stimuli requiring the same or
different response. This makes it unnecessary to propose that positives
and negatives are classified by different perceptual strategies. Instead we
can say that all the perceptual processing is done alike, by automatic
feature analysis yielding combinational representations, and that the relative response times to a large extent reflect processes of activation and
interference following full perceptual analysis. This account in no way
preempts the operation of other factors discussed by Nickerson (1978). It
does cast doubt, though, on the relevance of these types of study to
perceptual processes.
Much the same kind of account can be given of evidence from visual
search (Brand, 1971; Ingling, 1972) that the category of a character can be
known before its identity. Suppose that the effect of instructions to search
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
289
for a category is to associate a response with all members of that category
(as we have assumed for classification of multidimensional stimuli). Suppose also that all items within the effective visual field are automatically
fully identified nonconsciously. The response to a target item, of orientation or exclamation, will be activated, without the subject knowing which
it was. The identification of an item (automatic access to its internal
representation) causes activation of associated items. If all items in a
category are associated with the same response, less response interference will arise from items in the visual field than when only one item from
that category is so associated, especially if members of one category are
stored together and separately from different categories. Thus search
specifications may well affect responses rather than perception. That we
may consciously know higher level aspects before specific characteristics
of a target, even that a target has been found before its location, can be
attributed to principles of recovery discussed earlier.
What of Jonides and Gleitman’s (1972) finding that exactly the same
stimulus, an “O”, is found more efficiently among letters when it is called
a “zero” than when an “oh,” and vice versa for a context of numerals?
The solution is to invoke exactly the same argument as applied to other
classification experiments. Whichever the target is called, the response
will be associated with the “lexical” representation of the symbol, which
is accessed automatically. The lexical representations of numerals are
stored close to one another but not with letters and vice versa. Thus
although the symbol “0” when encountered will access both the lexical
representation of “oh” and “zero,” only one will have a response tag. It
will also associatively activate other members of each set of lexical representations (letters and digits). But whichever set is being used as a context
for the search will have negative or “no response” instructions attached
to them (see Marcel, 1977), if only to avoid false hits. Thus when the
symbol is called “oh” and is presented among letters, inhibition and
response competition (respond vs withhold any response) will be generated by contextual items. When the response is associated with “zero”
among a context of letters, inhibition cannot be passed on to the lexical
entry tagged for response, since it is in a lexical set unassociated with the
context items.
Although the above arguments may seem tortuous and are certainly not
watertight, they are meant to offer the possibility of an alternative view to
that motivated by the Identity Assumption. The usual view is that response characteristics in perceptual tasks reflect the perceptual processes
nominally addressed by the task. The alternative view is that the response
characteristics reflect the results of other perceptual processes (the acknowledged basis of Stroop effects) or response processes themselves. If
we accept that unconscious processes precede conscious ones, then at
290
ANTHONY
J. MARCEL
least it must be admitted that much of what we know so far about pattern
recognition is related to those processes which convert nonconscious
representations into conscious ones or to conscious processes themselves. Our methods, relying on “direct” tests and thus based on phenomenal percepts have afforded us little empirical insight into the initial,
perhaps fundamental, perceptual processes themselves. Physiology is of
little help to functional analyses. Artificial Intelligence attempts at Image
Formation and Scene Analysis, although supposedly addressed to these
problems, until recently (Mat-r, 1976; Marr & Nishihara, 1978a, 1978b)
have been also of restricted help. To echo Turvey (1974), “The Hoffding
Step (Hoffding, 1891) remains very much a mystery.”
CONCLUDING
REMARKS
An approach has been presented to the relations between nonconscious
perceptual processes and conscious experience. This approach by substituting paradigm assumptions, attempts to account for diverse
phenomena in a reasonably unified fashion. While many of the constructs
involved and the accounts of specific phenomena may differ little from
existing constructs and accounts, the roles and emphases given to them in
the general scheme are quite different.
The separation enjoined in this paper raises several issues. The most
immediate concerns methodology. The experiments in the preceding
paper themselves provoked the suggestion that if one wants to know
about those mechanisms which actually analyze and redescribe sensory
data and which largely support our experience and actions, then one
needs to use indirect rather than direct techniques of assessment. The
present theoretical approach suggests other cautions. Reports of phenomenological experience or responses based on it will tell us about conscious
percepts, but will not only be molded by our tacit belief systems (Nisbett
& Wilson, 1977), they will be affected by nonconscious processes. They
will also actively impede the revelation of the process of concern. Further,
since it has been suggested that conscious and nonconscious representations differ qualitatively and have different effects, it may be wise even to
preclude conscious awareness of the stimuli of concern by masking or
equivalent procedures.
Secondly, rather more attention could be devoted to consciousness per
se. One reason why what has been nonconsciously perceived is not revealed by direct report, is that people attempt to base their behavior on
some.notion of “rationality.”
One does not report whatever comes into
one’s mind if one’s task is to report what has been presented and one has
no conscious visual or auditory impression. Now this leads to an interesting speculation. It might well be that both the nature of our experience and what we acknowledge as a percept are subject to our tacit belief
CONSCIOUS
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
291
systems and culturally transmitted assumptions. It has been suggested
here that conscious experience is an account of evidence in terms of
available perceptual hypotheses. In this sense phenomenal experience is
an attribution. But attributions must have a basis for inference in a set of
rules or beliefs. Our accounts tend to consistency and particular types of
representations of the world (an articulated world, effable categories of
experience, attribution of percepts or sensations to a source). Perhaps
experience does not need to be of this nature but is so in our culture
because of epistemological presuppositions. These presuppositions constitute part of what Polanyi (1964) described as tacit knowledge. Particularly good analyses of them are supplied by Abelson (1968) in the domain
of implicit causal theories, and by authors discussed by Goffman (1975) in
the domain of social cognition.
Two phenomena are particularly relevant to the application of these
concepts to the principal concerns of this paper. First, in a recent study
(Rapoport, Buchsbaum, Weingartner, Zahn, Ludlow, Mikkelsen, Langer,
& Bunney, 1980) the subjective effects of d-amphetamine were compared in normal children and adults and hyperactive children. Results for
adults confirmed previous findings of feelings of euphoria. However, despite physiological measures, in neither group of children was this or any
other particular feeling clearly observed, except for some reports of “feel
funny, not like myself.” Robbins and Sahakian (1979) postulate that the
lack of what might be assumed to be an automatic effect may in fact be
due to a lack of experience in labeling emotional states: quite apart from
verbalization, the child cannot have a phenomenal experience unless he
or she has a construct for it. Thus it is a real possibility that autonomic
processes can have no issue in phenomenal experience or in modulating
even the tone of behavior unless an appropriate cognitive, phenomenological category has been acquired. That is, until a category has been acquired, processing effects cannot be parsed in terms of it either nonconsciously or consciously.
The second phenomenon concerns Weiskrantz et al.‘s (1974)
“blindsight”
patient. When he successfully made discriminations of
orientation and shape in the phenomenally blind field, they were not
based on visual experience. When asked about the shape discrimination
(0 vs X) he said he saw nothing but had a “feeling” that it was “smooth”
or “jagged.” The crucial point is that after being surprised by video
recordings of his own performance and reading reports of the studies,
Weiskrantz reports that his patient’s consciously blind field as measured
by perimetry has shrunk. Zihl (1981) has reported other cases where
phenomenal detection, acuity, and color identification has been recovered
in part of the previously blind field after practice at locating light sources.
While Zihl speculates that this is due to cortical recovery, there is no
292
ANTHONY
J. MARCEL
evidence for it, and in some of the cases it seems neurologically quite
improbable. Is it possible that the patient’s criteria for a percept have
altered? The purpose of these citations is to hint at the role of cognition in
determining the very kind of experience which has long been assumed by
psychophysicists to be nonmediated.
In conclusion, while the approach outlined here is highly speculative
and remains conceptually loose and underspecified, it appears to have
fairly far-reaching implications and it is hoped that it will provoke at least
some reconsiderations among cognitive psychologists.
APPENDIX
Extracts from interview with patient describing some of his visual difficulties.4
Interviewer: I think you also had some difficulty in perceiving people as
units.
Patient: Oh, oh yes. There was a slightly different effect I think, that if I
saw a complex object, such as a person, and there were several people
in my field of view, I sometimes saw the different parts of the people
as not, in a sense, belonging together, although . . . if a given person
moved so that all the parts of him went in one direction, that would . . .
tend to make him into a single object. Otherwise there was this confusion of lots of things, all of which were there, but did not seem to
belong together. . . . Several of these cases of things not belonging
together gave quite absurd results. For instance, I do remember one
case where there was what seemed to me to be one object which was
partly motor car, partly tree and partly a man in a cricket shirt. They
seemed somehow to belong together. More frequently, however, a lot
of things which to any ordinary viewer would be parts of the same
thing were parts of different things.
I: So it was essentially common movement that created these units.
P: Yes. . . . I think that was perhaps the most frightening case. A common
color, especially in the case of clothes . . . when there were crowds of
people together for instance on the lawn or on the beach, also formed a
unifying thing. . . . The effect was much more striking when a large
number of objects were on the same table . . . it was not obvious what
belonged to what; there were a whole lot of different things and in fact
sometimes they-when one only saw a small object one could hardly
say anything more than one saw a colored patch. . . . (if somebody I
knew was speaking to me) . . . it sounds quite absurd but there were
two distinct things. One was that so and so was speaking to me and I
4 I am grateful to Professor 0. L. Zangwill for providing
interview and for allowing me to quote from it.
me with the transcript
of this
CONSCIOUS
I:
P:
I:
P:
I:
P:
EXPERIENCE
AND
PERCEPTUAL
PROCESSES
293
could hear and understand what he said; two, that he was standing in
front of me and I could see his mouth moving, but I noticed that the
mouth moving did not belong to what I heard any more than a -than
one of the old talkie pictures would make sense if the voice tape had
been the wrong tape for the conversation. That was absolutely quite
fantastically exiciting. . . .
Was this a failure to localise the source of the voice?
No. No. It was as though they were two different things.
They didn’t belong together.
Didn’t belong together at all. Another example of this-one normally
sees the world being made up of things. . . .
If I could go back to pictures for a moment. If you looked at a picture,
say in the daily paper, could you see at once what it was of?
Oh no. I should add that very often looking at pictures in daily papers
for instance, especially if they were rather large pictures, I had to look
at different bits and found it very difficult to combine them into a
whole. This was not due to the poor quality of the print. . . . It did not
usually apply to portraits of single people, but if there were a number
of people in the picture at the same time it was very difficult to say
how many they were or what they were doing. . . . Sometimes it
seemed as though I had moved my eyes unintentionally because one
object would be replaced by something which in fact could be seen in
another part of the visual field. . . . (discussing his present reading
ability) I ought to add that even reading along a line of closely typed
print . . . I very often jump from one line to another without realising
that I have done so-so that I read quite convincingly the first half of
one line and it seems to follow on directly, be followed directly by the
second half of a compietely other line resulting in complete nonsense.
REFERENCES
Abelson, R. P. Psychological implication. In R. P. Abelson, E. Aronson, & W. J. McGuire
(Eds.), Theories of cognifive consistency: A sourcebook. Chicago: Rand McNally, 1968.
Adler, A. Disintegration and restoration of optic recognition in visual agnosia. Archives of
Neurology and Psychiatry,
1944, 51, 243-259.
Allport, D. A. On knowing the meaning of words we are unable to report: The effects of
visual masking. In S. Domic (Ed.), Attenfion and performance VI. New York/London:
Academic Press, 1977.
Averbach, E., & Coriell, A. S. Short-term memory in vision. Bell Systems Technical Journal, 1961, 40, 309-328.
Balint, R. Seelenlahmung des “Schauens,”
optische Ataxie, raumliche Storung der Aufmerksamkeit. Monatsschrift fur Pqychiatrie und Neurologie,
1909, 2.5, 57-71.
Battista, J. R. The science of consciousness. In K. S. Pope & J. L. Singer (Eds.), The
stream of consciousness. New York: Plenum, 1978.
Bender, M. B. Disorders in perception. Springfield, IL: Thomas, 1952.
Biederman, I. Perceiving real-world scenes. Science, 1972, 177, 77-79.
294
ANTHONY
J. MARCEL
Blumstein, S. E., Milberg, W., and Shrier, R. Semantic processing in aphasia: Evidence
from an auditory lexical decision task. Brain and Language, 1982, 17, 301-315.
Bouma, H. Interaction effects in parafoveal letter recognition. Nature (London), 1970, 226,
177- 178.
Bradshaw, J. L. Peripherally presented and unreported words may bias the perceived
meaning of a centrally fixated homograph. Journal ofExperimental
Psychology, 1974,
103, 1200- 1202.
Brain, Sir R. Speech disorders. London: Butterworths, 1961.
Brand, J. Classification without identification in visual search. Quarterly Journal ofExperimental Psychology,
1971, 23, 178- 186.
Bregman, A. S., & Rudnicky, A. I. Auditory segregation: Stream or streams? Journal of
Experimental Psychology: Human Perception and Performance,
1975, 1, 263-267.
Breitmeyer, B. G., & Ganz, L. Implications of sustained and transient channels for theories
of visual pattern masking, saccadic suppression and information processing.
Psychological
Review, 1976, 83, l-36.
Brentano, F. C. Psychologie von dem Empirischen Standpunkt. Leipzig: Duncker and
Humblot, 1874. Trans. from 1924 ed. by A. C. Rancurello, D. B. Terrell, & L. L.
McAlister. L. L. McAlister (Ed.), London: Routledge & Kegan Paul, 1973.
Bridgeman, B. Metacontrast and lateral inhibition. Psychological
Review,
1971, 78,
528-539.
Butterworth, B. Hesitation and the production of verbal paraphasias and neologisms in
jargon aphasia. Brain and Language, 1979, 8, 133-161.
Campbell, R. N. Language acquisition, psychological dualism and the definition of pragmatics. (Paper delivered at conference on possibilities and limitations of pragmatics,
Urbino, July 1979.) In J. Verschueren & H. Parret (Eds.), Possibilities and limitations
of pragmatics. Amsterdam: John Benjamins, 1980.
Carr, T. H., & Bacharach, V. R. Perceptual tuning and conscious attention: Systems of
input regulation in visual information processing. Cognition, 1976, 4, 281-302.
Clowes, M. B. On seeing things. Artificial Intelligence, 1971, 2, 79-116.
Collins, A. M., & Loftus, E. F. A spreading-activation theory of semantic processing.
Psychological
Review, 1975, 82, 407-428.
Coltheart, M. Iconic memory and visible persistence. Perception and Psychophysics, 1980,
27, 183-228.
Corteen, R. S., & Wood, B. Autonomic responses to shock-associated words in an unattended channel. Journal of Experimental Psychology, 1972, 94, 308-313.
Cutler, A., & Norris, D. Monitoring sentence comprehension. In W. E. Cooper & E. C. T.
Walker (Eds.), Sentence processing: Studies in honour of Merrill Garrett. Hillsdale,
NJ: Erlbaum, 1978.
Cutting, J. E. Auditory and linguistic processes in speech perception: Inferences from six
fusions in dichotic listening. Psychological Review, 1975, 83, 114- 140.
Cutting, J. E. Six tenets for event perception. Cognition, 1981, 10, 71-78.
Dember, W. N., & Purcell, D. G. Recovery of masked visual targets by inhibition of the
masking stimulus. Science, 1967, 157, 1335- 1336.
Dember, W. N., Schwartz, M., & Kocak, M. Substantial recovery of a masked visual target
and its theoretical interpretation. Bulletin of the Psychonomic Society, 1978, 11,
285-287.
Deutsch, J. A., & Deutsch, D. Attention: Some theoretical considerations. Psychological
Review, 1963, 70, 80-90.
Dixon, N. F. Subliminal perception: the nature of a controversy. London: McGraw-Hill,
1971.
Donaldson, M. Children’s minds. Glasgow: Fontana, 1978.
CONSCIOUS EXPERIENCE
AND PERCEPTUAL PROCESSES
295
Fischler, I. Associative facilitation without expectancy in a lexical decision task. Journal of
Experimental
Psychology:
Human Perception
and Performance,
1977, 3, 18-26.
Foss, D. J., & Swinney, D. A. On the pyschological reality of the phoneme: Perception,
identification and consciousness. Journal of Verbal Learning and Verbal Behavior,
1973, 12, 246-257.
Fowler, C. A., Wolford, G., Slade, R., & Tassinary, L. Lexical access with and without
awareness. Journal of Experimental Psychology: General, 1981, 110, 341-362.
Fromkin, V. A. (Ed.) Speech errors as linguistic evidence. The Hague: Mouton, 1973.
Goffman, E. Frame analysis. Harmondsworth: Penguin, 1975.
Goldstein, K. Language and language disturbances. New York: Grune & Stratton, 1948.
Gregory, R. L. The intelligent eye. London: Weidenfeld & Nicholson, 1970.
Grindley, G. C., & Townsend, V. Voluntary attention in peripheral vision and its effects on
acuity and differential thresholds. Quarterly Journal of Experimental Psychology, 1968,
20, 11-19.
Haber, R. N. The nature of the effect of set on perception. Psychological Review, 1966,73,
335-351.
Paris: Librairie Larousse, 1972.
Hecaen, H. Introduction d la neuropsychologie.
Hecaen, H., & Albert, M. L. Human neuropsychology.
New York: Wiley, 1978.
Henderson, L. Visual and verbal codes: Spatial information survives the icon. Quarterly
Journal
of Experimental
Psychology,
1972, 24, 439-447.
Hochberg, J. Attention, organization and consciousness. In D. I. Mostofsky (Ed.), Attention: Contemporary theory and analysis. New York: Appleton-Century-Crofts,
1970.
Hoffding, H. Outlines of psychology. New York: Macmillan, 1891.
Humphrey, N. K. What the frog’s eye tells the monkey’s brain. Brain, Behavior and Evolution, 1970, 2, 324, 337.
Humphrey, N. K. Vision in a monkey without striate cortex: A case study. Perception,
1974, 3, 241-255.
Husserl, E. Cartesian meditations, 1929. Trans. D. Cairns. The Hague: Martinus Nijhoff,
1960.
lngling, N. Categorization: a mechanism for rapid information processing. Journal of Experimental Psychology,
1972, 94, 239-243.
Jacobson, J. Z. Effects of association upon masking and reading latency. Canadian Journal
of Psychology,
1973, 27, 58-69.
Jacobson, J. Z. Interaction of similarity to words of visual masks and targets. Journal of
1974, 102, 431-434.
Experimental Psychology,
Jacobson, J. Z., & Rhinelander, G. Geometric and semantic similarity in visual masking.
Journal of Experimental
Psychology: Human Perception and Performance,
1978, 4,
224-231.
Johansson, G. Projective transformations as determining visual space perception. In R. B.
MacLeod & H. L. Pick, Jr. (Eds.), Perception: Essays in honor of James J. Gibson.
Ithaca, NY: Cornell Univ. Press, 1974.
Johnston, J. C., & McClelland, J. L. Visual factors in word perception. Perception and
Psychophysics,
1973, 14, 365-370.
Jonides, J., & Gleitman, H. A conceptual category effect in visual search: “0” as letter or
digit. Perception and Psychophysics, 1972, 10, 457-460.
Kinsbourne, M., & Warrington, E. K. A disorder of simultaneous form perception. Brain,
1962, 85, 461-486.
Kinsbourne, M., & Warrington, E. K. The localizing significance of limited simultaneous
visual form perception. Brain, 1963, 86, 697-702.
Klatt, D. H. Speech perception: a model of acoustic-phonetic analysis and lexical access.
Journal of Phonetics, 1979, 7, 279-312.
Koffka, K. Principles of gestalt psychology. New York: Harcourt, Brace, 1935.
296
ANTHONY
J. MARCEL
Kolers, P. A. Intensity and contour effects in visual masking. Vision Research, 1962, 2,
277-294.
Kolers, P. A. Some pyschological aspects of pattern recognition. In P. A. Kolers & M. Eden
(Eds.), Recognizing patterns. Cambridge: MIT Press, 1968.
Kolers, P. A., & Rosner, B. S. On visual masking (metacontrast): Dichoptic observations.
American Journal of Psychology,
1960, 73, 2-21.
Kornfeld, J. R. Implications of studying reduced consonant clusters in normal and abnormal
child speech. In R. N. Campbell & P. J. Smith (Eds.), Recent advances in the study of
language: Language development and mother-child
interaction.
NATO Conference
Series, III 4A, New York: Plenum, 1977. pp. 413-424.
Kristofferson, A. B., Galloway, J., & Hanson, R. G. Complete recovery of a masked visual
target. Bulletin of the Psychonomic Society, 1979, 13, 5-6.
Lashley, K. S. Cerebral organization and behavior. In The brain and human behavior,
proceedings of the association for research on nervous and mental disease. Baltimore:
Williams & Wilkins, 1958.
Lehmkuhle, S., & Fox, R. Effect of depth separation on metacontrast masking. Journal of
1980, 6, 605-621.
Experimental Psychology: Human Perception and Performance.
Lewis, J. L. Semantic processing of unattended messages under dichotic listening. Journal
of Experimental
Psychology,
1970, 85, 225-228.
Liberman, A. M. The grammars of speech and language. Cognitive Psychology, 1970, 1,
301-323.
Liepmann, H. Uber die agnostischen Storungen. Neurologisches Zentralblutt,
1908, 27,
609-617.
Lissauer, H. Ein Fall von Seelenblindheit nebst einem Beitrag zur Theorie derselben. Archiv. fiir Psychiatric
und Nervenkrankheiten,
1890, 21, 222-299.
Luria, A. R. Traumatic aphasia. Moscow: Academy of Medical Science of the USSR, 1947.
Luria, A. R. Psycho-physiological
features of writing. Moscow: Academy of Psychological
Science of the Russian Soviet Federative Socialist Republic, 1950.
Luria, A. R. Disorders of “simultaneous perception” in a case of bilateral occipito-parietal
brain injury. Brain, 19.59,82, 437-449.
McCauley, C., Parmelee, C., Sperber, R., & Carr, T. Early extraction of meaning from
pictures and its relation to conscious identification. Journal of Experimental Psychology: Human Perception
and Performance,
1980, 6, 265-276.
Mackworth, N. H., & Bruner, J. S. How adults and children search and recognize pictures.
Human Development,
1970, 13, 149- 177.
Mandler, G. Consciousness: Respectable, useful and probably necessary. In R. Solso (Ed.),
Information processing and cognition: The Loyola symposium. Hillsdale, NJ: Erlbaum,
1975.
Marcel, A. J. Some constraints on sequential and parallel processing, and the limits of
attention. In A. F. Sanders (Ed.), Attention and performance.
Amsterdam: NorthHolland, 1970. Vol. 3.
Marcel, A. J. Negative set effects in character classification: A response retrieval view of
reaction time. Quarterly Journal of Experimental Psychology, 1977, 29, 31-48.
Marcel, A. J. Phonological awareness and phonological representation-Investigation
of a
specific spelling problem. In U. Frith (Ed.), Cognitive processes in spelling. New
York/London: Academic Press, 1979.
Marcel, A. J. Conscious and preconscious recognition of polysemous words: Locating the
selective effects of prior verbal context. In R. S. Nickerson (Ed.), Attention and performance, VIII. Hillsdale, NJ: Erlbaum, 1980.
Marcel, A. J. Conscious and unconscious perception: Experiments on visual masking and
word recognition. Cognitive Psychology, 1983, 15, 197-237.
CONSCIOUS EXPERIENCE
AND PERCEPTUAL PROCESSES
297
Marcel, A. J., & Patterson, K. E. Word recognition and production: Reciprocity in clinical
and normal studies. In J. Requin (Ed.), Attention and performance VII, Hillsdale, NJ:
Erlbaum, 1978.
Marcel, A. J., & Patterson, K. E. An aphasic reading impairment in normal people: The
hemispheric andprocessing
locus of word-class effects. Paper delivered to the Experimental Psychology Society, Reading, July 1976.
Marcus, S. M. ERIS-Context sensitive coding in speech perception. Manuscript 367, Institute for Perceptual Research, Eindhoven, Holland, December 1979.
Marr, D. Early processing of visual information. Philosophical Transactions of the Royal
Society, B, 1976, 275, 483-524.
Mat-r, D., & Nishihara, H. K. Visual information processing: Artificial intelligence and the
sensorium of sight. Technology Review, 1978, 81(l) MIT. (a)
Marr, D., & Nishihara, H. K. Representation and recognition of the spatial organization of
3-dimensional shapes. Proceedings of the Royal Society of London, Series B. 1978,200,
269-294. (b)
Massaro, D. W., & Oden, G. C. Speech perception: A framework for research and theory.
In N. J. Lass (Ed.), Speech and Language: Advances in basic research and pracfice.
New York: Academic Press, 1980.
Mattingly, I. G. Reading, the linguistic process and linguistic awareness. ln J. F. Kavanagh
& J. G. Mattingly (Eds.), Language by ear and by eye. Cambridge, MA: MIT Press,
1972.
Merikle, P. M., Coltheart, J., & Lowe, D. G. On the selective effects of a patterned masking
stimulus. Canadian Journal of Psychology, 1971, 25, 264279.
Mewhort, D. J. K. Familiarity of letter sequences, response uncertainty, and the tachistoscopic recognition experiment. Canadian Journal of Psychology, 1967, 21, 309-321.
Meyer, D. E., Schvaneveldt, R. W., & Ruddy, M. G. Loci of contextual effects on visual
word recognition. In P. M. A. Rabbit & S. Dornic (Eds.), Attenfion undperformance,
V. New York: Academic Press, 1975.
Milberg, W., & Blumstein, S. E. Lexical decision and aphasia: Evidence for semantic
processing. Brain and Language, 1981, 14, 371-385.
Miller, G. A. Psychology: The science of mentnl life. New York: Harper & Row, 1962.
Minsky, M. A framework for representing knowledge. In P. A. Winston (Ed.), The psychology of computer vision. New York: McGraw-Hill, 1975.
Moore, J., & Newell, A. How can MERLIN understand? In L. W. Gregg (Ed.), Knowledge
and cognition. London: Erlbaum, 1974.
Morais, J., Cary , L., Alegria, J., & Bertelson, P. Does awareness of speech as a sequence of
phones arise spontaneously? Cognition, 1979, 7, 323-33.
Morton, J. Grammar and computation in language behaviour. In J. C. Catford (Ed.), Studies
in language and language behaviour. Center for Research in language and language
behavior, Progress Report No. VI. Ann Arbor: University of Michigan, 1968.
Neely, J. M. Semantic priming and retrieval from lexical memory: Roles of inhibitionless
spreading activation and limited capacity attention. Journal of Experimental Psychology: General, 1977, 106, 226-254.
Neisser, U. Visual imagery as process and as experience. In J. S. Antrobus (Ed.), Cognition
and affect. Boston: Little, Brown, 1970.
Nickerson, R. S. Binary classification reaction time: A review of some studies of human
information processing capabihties. Psychonomic Monograph Supplements, 1972,
4(17), 275-318.
Nickerson, R. S. On the time it takes to tell things apart. In J. Requin (Ed.), Atfention and
performance,
VIII. Hillsdale, NJ: Erlbaum, 1978.
298
ANTHONY
J. MARCEL
Nisbett, R. E., & Wilson, T. Dec. Telling more than we know: Verbal reports on mental
processes. Psychological Review, 1977, 84, 231-259.
Oden, G. C., & Massaro, D. W. Integration of featural information in speech perception.
Psychological
Review, 1978, 85, 172-191.
Paivio, A. Mental imagery in associative learning and memory. Psychological Review, 1969,
76, 241-263.
Palmer, S. E. Hierarchical structure in perceptual representation. Cognitive Psychology,
1977, 9, 441-474.
Piaget, J., & Inhelder, B. The psychology of the child. New York: Basic Books, 1969.
Polanyi, M. Personal knowledge: Towards a post-critical philosophy. New York: Harper,
1964.
Polanyi, M. The tacit dimension. Garden City, NY: Doubleday, 1966.
Poppel, E., Held, R., & Frost, D. Residual function after brain wounds involving the central
visual pathways in man. Nature (London), 1973, 243, 295-296.
Posner, M. I. Orienting of attention (Text of the 7th Sir Frederic Bartlett lecture). Quarterly
Journal of Experimental Psychology,
1980, 32, 3-25.
Posner, M. I., & Klein, R. M. On the functions of consciousness. In S. Kornblum (Ed.),
Attention and performance,
IV. New York: Academic Press, 1973.
Posner, M. I., & Snyder, C. R. R. Attention and cognitive control. In R. L. Solso (Ed.),
Information processing and cognition: The Loyola symposium. Hillsdale, NJ: Erlbaum,
1975.
Purcell, D. G., & Stewart, A. L. U-shaped backward masking functions with nonmetacontrast paradigms. Psychonomic Science, 1970, 21, 361-363.
Rapoport, J. S., Buchsbaum, M. S., Weingartner, H., Zahn, T. P., Ludlow, C., Mikkelsen,
E. J., Langer, D., & Bunney, W. E. Dexrro-amphetamine: Cognitive and behavioral
effects on normal and hyperactive children and normal men. Archives of General
1980, 37, 933-943.
Psychiatry,
Ratcliff, G., & Cowey, A. Disturbances of visual perception following cerebral lesions. In
D. J. Osborne, M. M. Gruneberg, & R. Eiser (Eds.), Psychology and medicine. New
York/London: Academic Press, 1979.
Reicher, G. M. Perceptual recognition as a function of stimulus material. Journal of Experi1969, 21, 275-280.
menral Psychology,
Robbins, T. W., & Sahakian, B. J. “Paradoxical” effects of psychomotor stimulant drugs in
hyperactive children from the standpoint of behavioural pharmacology. Neuropharmacology, 1979, l&931-950.
Rubens, A. B. Agnosia. In K. M. Heilman & E. Valenstein (Eds.), Clinical neuropsychology. New York: Oxford Univ. Press, 1979.
Rumelhart, D. E. Schemata: The building blocks of cognition. In R. Spiro, B. Bruce, &
W. Brewer (Eds.), Theorefical
issues in reading comprehension.
Hillsdale, NJ:
Erlbaum, 1978.
Schachter, S. Cognition and peripheralist-centralist controversies in motivation and emotion. In M. S. Gazzaniga & C. Blakemore (Eds.), Handbook of psychobiology.
New
York: Academic Press, 1975.
Schneider, G. E. Two visual systems. Science, 1967, 163, 895-902.
Schultz, D. W., & Eriksen, C. W. Do noise masks terminate target processing? Memory and
Cognition, 1977, 5, 90-96.
Selfridge, 0. Pandemonium: A paradigm for learning. In Mechanization
of thought processes. London: H.M. Stationery Office, 1959.
Shallice, T. Dual functions of consciousness. Psychological Review, 1972, 79, 383-393.
Shallice, T. The dominant action system: An information-processing approach to con-
CONSCIOUS EXPERIENCE
AND PERCEPTUAL PROCESSES
299
sciousness. In K. S. Pope & J. L. Singer (Eds.), The stream ofconsciousness:
Scientific investigations into the flow of human experience. New York: Plenum, 1978.
Shallice, T., & McGill, J. The origins of mixed errors. In J. Requin (Ed.), Atrenfion and
performance, VII. Hillsdale, NJ: Erlbaum, 1978.
Shaw, R., & Pittenger, J. Perceiving the face of change in changing faces: Implications for a
theory of object perception. In R. Shaw & J. Bransford (Eds.), Perceiving, acfing and
knowing. Hillsdale, NJ: Erlbaum, 1977.
Sperling, G. The information available in brief visual presentations. Psychological Monographs, 1960, 74, (1 I, Whole No. 498).
Sperling, G. A model for visual memory tasks. Human Facfors, 1963, 5, 19-31.
Squire, L. R. The neuropsychology of memory. Annual Review of the Neurosciences, 1982,
5, 241-273.
Strange, J. R. A Search for the sources of the stream of consciousness. In K. S. Pope &
J. L. Singer (Eds.), The stream of consciousness. New York: Plenum, 1978.
Studdert-Kennedy, M. The emergency of phonetic structure. Cognirion, 1981, 10,301-306.
Taylor, A. M., & Warrington, E. K. Visual agnosia: A single case report. Cortex, 1971, 7,
152- 161.
Taylor, Cl. A., & Chabot, R. J. Differential backward masking of words and letters by masks
of varying orthographic structure. Memory and Cognition, 1978, 6, 629-635.
Treisman, A. M. Binocular rivalry and stereoscopic depth perception. Quarterly Journal of
Experimental
Psychology,
1962, 14, 23-27.
Treisman, A. M., & Gelade, G. A feature-integration theory of attention. Cognitive Psychoiogy, 1980, 12, 197-236.
Treisman, A. M., Sykes, M., & Gelade, G. Selective attention and stimulus integration. In
S. Dornic (Ed.), Affe!ltion undperformance,
VI. New York/London: Academic Press,
1977.
Trevarthen, C. B. Two mechanisms of vision in primates. Psychologische Forschung, 1968,
31, 299-337.
Turvey, M. T. On peripheral and central processes in vision: Inferences from an
information-processing analysis of masking with patterned stimuli. Psycho/ogicaf Review, 1973, 80, l-52.
Turvey, M. T. Constructive theory, perceptual systems and tacit knowledge. In W. B.
Weimer & D. S. Palermo (Eds.), Cognition and the symbolic processes. Hillsdale, NJ:
Erlbaum, 1974.
Turvey, M. T. Preliminaries to a theory of action with reference to vision. In R. Shaw &
J. Bransford (Eds.), Perceiving, acting and knowing: Toward an ecological psychology.
Hillsdale, NJ: Erlbaum, 1977. (a)
Turvey, M. T. Contrasting orientations to the theory of visual information processing.
Psychological
Review, 1977, 84, 67-88. (b)
Turvey, M. T., & Fertig, J. Polarity on the semantic differential and release from proactive
interference in short-term memory. Journal of Verbal Learning and Verbal Behavior.
1970, 9, 439-443.
Turvey, M. T., Fertig, J., & Kravetz, S. Connotative classification and proactive interference in short-term memory. Psychonomic Science. 1969, 16, 223-24.
Tweedy, J. R., Lapinski, R. H., & Schvaneveldt, R. W. Semantic-context effects on word
recognition: Influence of varying the proportion of items presented in an appropriate
context. Memory and Cognition, 1977, 5, 84-89.
Tyler, H. R. Abnormalities of perception with defective eye movements (Balint’s syndrome). Cortex, 1968, 4, 154- 171.
Underwood, G. Semantic interference from unattended printed words. British Journal of
Psychology,
1976, 67, 327-328.
300
ANTHONY
J. MARCEL,
Underwood, G. Attention and the nonselective lexical access of ambiguous words. Canadian Journal of Psychology, 1980, 34, 72-76.
Volpe, B. T., Ledoux, J. E., & Gazzaniga, M. S. Information processing of visual stimuli in
an “extinguished” field. Nature (London), 1979, 282, 722-724.
von Wright, J. M. Selection in visual immediate memory. Quarterly Journal ofExperimental
Psychology,
1968, 20, 62-68.
von Wright, J. M. On the problem of selection in iconic memory. Scandinavian Journal of
Psychology, 1972, 13, 159-171.
Walker, P. Binocular rivalry: Central or peripheral selective processes. Psychological
Bulletin, 1978, 85, 376-389.
Warren, R. M. Auditory illusions and perceptual processes. In N. J. Lass (Ed.), Contemporary issues in experimental phonetics. New York: Academic Press, 1976.
Warrington, E. K. The selective impairment of semantic memory. Quarterly Journal of
Experimental Psychology,
1975, 27, 635-657.
Weiskrantz, L. Trying to bridge some neuropsychological gaps between monkey and man.
British Journal of Psychology,
1977, 68, 431-445.
Weiskrantz, L., Warrington, E. K., Sanders, M. D., & Marshall, J. Visual capacity of the
hemianopic field following a restricted occipital ablation. Brain, 1974, 97, 709-728.
Weisstein, N. W-shaped and U-shaped functions obtained for monoptic and dichoptic diskdisk masking. Perception and Psychophysics, 1971, 9, 275-278.
Weisstein, N., & Harris, C. S. Visual detection of line segments: An objects-superiority
effect. Science, 1974, 186, 752-755.
Wheeler, D. D. Processes in word recognition. Cognitive Psychology, 1970, 1, 59-85.
Wickelgren, W. A. Phonetic coding and serial order. In E. C. Carterette & M. P. Friedman
(Eds.), Handbook of Perception. New York: Academic Press, 1976. pp. 227-264,
Vol. 7.
Wickens, D. D. Encoding categories of words: An empirical approach to meaning.
Psychological Review, 1970, 77, 1- 15.
Wickens, D. D., & Engle, R. W. Imagery and abstraction in short-term memory. Journal
of Experimental Psychology,
1970, 84, 268-272.
Williams, A., & Weisstein, N. Line segments are perceived better in a coherent context than
alone: An object-line effect in visual perception. Memory and Cognition, 1978, 6,
85-90.
Willows, D. M., & MacKinnon, G. E. Selective reading: Attention to the “unattended”
lines. Canadian Journal of Psychology, 1973, 37, 292-304.
Wolpert, I. Die Simultanagnosie. Zeitschrift fir Gesammter Neurologie und Psychiatric,
1924, 93.
Womersley, M. A contextual effect in feature detection with application of signal detection
methodology. Perception and Psychophysics, 1977. 21, 88-92.
Yarbus, A. Eye movements in vision. New York: Plenum, 1968.
Zihl, J. Recovery of visual functions in patients with cerebral blindness. Experimental Brain
Research, 1981, 44, 159- 169.
REFERENCE
NOTES
1. Creighton, P. Word recognition, masking and consciousness. Manuscript to be submitted as Ph.D. thesis to University of Cambridge.
2. Marcel, A. J. Is cortical blindness a problem of visual consciousness or visualfunction?
Paper presented at Fifth International Neuropsychology Society European Conference, Deauville, France, June 1982.
(Accepted December 1, 1982)