Tel Aviv University
Faculty of Social Sciences
Department of Psychology
The Effects of Past and Present Experience with Target and Context
on Visual Search Performance:
A New Approach to Latent Inhibition
Dissertation submitted for the degree of Doctor of Philosophy
The present research was completed under the supervision of
Professor R.E. Lubow
Submitted to the Senate of Tel Aviv University
Latent inhibition (LI) is a phenomenon in which a stimulus that was registered as irrelevant for the organism in the past enters into new associations with difficulty as compared to a novel stimulus. Human LI experiments have used three main paradigms, classical conditioning, instrumental conditioning, and associative rule learning. All of these, with but a few minor exceptions, have employed a between-subject design, in which one group is preexposed in a context where the to-be-tested stimulus is irrelevant, and then exhibits retarded learning in the test phase (PE group), and another group is not preexposed and manifests normal learning in the test phase (NPE group). The present series of studies employs a fourth paradigm to elicit LI. It is based on a visual search procedure and it uses both a between and within-subject designs. The new approach has several advantages over older LI methods. First, the visual search procedure measures attention more directly than traditional LI procedures. As most explanations of LI are based on attentional processes, which occur during preexposure and test, such a new approach is required. Second, the ability to measure LI with a within-subject design has advantages, especially in LI studies with special populations (e.g., schizophrenics, Parkinson patients, ADHD children etc.).
The new procedure is based on a visual search task in which a subject in the preexposure phase is repeatedly required to detect the presence or absence of a target among homogenous distractors (both distractors and target are nonsense shapes). In a subsequent test, the subject engages in the same task, only now the distractors/target presentation is heterogeneous. On a given trial, the distractors can be familiar, either the preexposure target, the preexposure distractor, or not familiar (novel). In the same way, the target can be familiar, either the preexposure target, or preexposure distractor, or not familiar. The different distractors familiarity condition and target familiarity conditions in test generate seven familiarity combinations in test. Two of these familiarity combinations were used to assess LI. When the distractor in test was familiar as a result of having been the target in preexposure, and the target in test was familiar as a result of having been the preexposure distractors, this re-creates the equivalent of the PE condition in traditional LI studies (relatively longer RTs). When, in the same test, the target was novel, this re-creates the equivalent of the NPE condition of traditional LI (relatively shorter RTs). Therefore, the differences in RTs between PE and NPE conditions in the visual search procedure can serve to index LI magnitude.
Experiment 1 describes the basic procedure in which all-possible test familiarity conditions were used. The results demonstrated the expected LI effect, as well as a novel pop-out effect (a novel stimulus on a familiar background is detected faster than a novel stimulus on a novel background).
In Experiment 2, in order to assess LI more efficiently, the test familiarity conditions were reduced from seven to four (2×2). The test target could be novel or familiar (preexposed distractor), the test distractors could be novel or familiar (preexposed target). Surprisingly, the LI effect was abolished, perhaps because of the omission of the test condition “target familiar as preexposure target” and “distractors familiar as preexposure distractors” (i.e., the identical condition to that of preexposure, and therefore labeled as a “reminder” trial). A review of the literature suggested that the reminder-trial may have served as a retrieval cue in Experiment 1, and also preserved context constancy from preexposure to test (of importance in traditional LI studies); it was hypothesized that the omission of the reminder trial in Experiment 2 resulted in the abolition of LI.
This issue is also related to a theoretical debates concerning LI. The main controversy revolved around two explanations for LI. One is represented by Conditioned Attention Theory and the other by Retrieval Failure Theory. The former claims that preexposure results in an association deficit which is not reversible. The latter claims that preexposure results in retrieval failure of memories encoded during that phase, an effect that is reversible. The results of Experiment 1 and 2 lend support to this theory.
Four additional experiments were conducted to address these issues. The experiments manipulated test familiarity conditions, test complexity (homogenous vs. heterogeneous), reminder cues, and contextual cues. It was demonstrated that context change from preexposure to test reduced LI in the visual search procedure, as in traditional LI studies. The most salient results were related to the reminder effect. Reminder cues were critical for eliciting LI in a heterogeneous but not in a homogenous test environment (defined by number of different trial-types in the test). It was suggested that failure of researchers in the past to design an effective within-subject LI procedure was a result of the absence of reminder trials.
In general, the results supported Retrieval Failure Theory. An attempt was made to integrate Conditioned Attention and Retrieval Failure Theories, as well as to have those theories accommodate other cognitive phenomena, particularly novel popout and negative priming.
When a stimulus is preexposed as irrelevant for the organism, namely when it has no consequences, it becomes difficult for that stimulus to enter into new associations. Any learning in which this stimulus will be involved will be less effective compared to learning with a novel stimulus. This phenomenon, latent inhibition (LI, e.g., Lubow & Moore, 1959; Lubow, 1989; Hall, 1991) was reported in many different species including humans. It was suggested that LI serves to protect the organism from the flooding of information from the environment by attenuating the processing of previously irrelevant stimuli. "Without latent inhibition, learning could be a cumbersome and inefficient process. Latent inhibition allows for the stimulus selectivity that is a requirement for rapid, efficient learning" (Lubow, 1989).
LI has been demonstrated in variety of mammalian species, including mouse, rat, rabbit, cat, sheep, goat, and humans, both young children and adults. It was investigated in special human populations, including schizophrenics, ADHD children, Parkinson's disease patients, anxiety-disordered children and adults. Furthermore, LI has been produced using many different procedures, including a number of variations of classical conditioning and instrumental conditioning, as well as some attentional tasks. LI is also found when the preexposed stimulus is taken from different modalities, including, auditory, visual, taste, and tactile. LI has been the subject of study within a variety of contextual issues in psychology, e.g. dopamine system (biological), learned helplessness (clinical), schizotypality (psychopathology), attention (cognitive), and even in consumer behavior (marketing). Since the term was first introduced (Lubow & Moore, 1959), LI has been widely investigated and found to be crucial in different learning tasks with animals and humans. Nevertheless, the theoretical basis of LI is still open to question.
The name "latent inhibition" is derived from the fact that the first study that demonstrated the effect (Lubow & Moore, 1959) was design to provide a classical conditioning analogue of "latent learning" and "sensory preconditioning". Both paradigms were prominent 50 years ago in the historical debates between different associative learning theories. One approach, represented mainly by Hull (1943), claimed that stimulus-response associations are reinforcement dependent. The opposing approach, represented mainly by Tolman (1932), claimed that associations could be generated as a result of mere proximity of stimuli, independent of reinforcement. Latent learning and sensory preconditioning use similar preexposure procedures. The main difference between them has to do with the type of stimulus. In latent learning, subjects are preexposed to an apparatus that later serves for instrumental conditioning (e.g., a maze). It was demonstrated that such preexposure facilitated subsequent learning in that maze. This learning is latent, in that the effects of non-reinforced preexposure become overt only in a subsequent test phase. In sensory preconditioning, there is also latent learning in preexposure phase, and an actualization of that learning in the subsequent test phase. Only, here, subjects are preexposed to a pair of neutral stimuli without any patent reinforcement. The association that is developed during preexposure is demonstrated in the subsequent classical conditioning learning phase. The first study of latent inhibition (Lubow & Moore, 1959) aimed to integrate the two approaches by combining the instrumental and classical conditioning procedures. As such, Lubow and Moore expected to find a facilitating effect as a result of stimulus preexposure. Such an effect was not realized. Instead, the reverse was obtained. The stimulus preexposure interfered with subsequent associations to that stimulus, thereby demonstrating what came to be called the “latent inhibition effect” (Lubow, 1989). The term "latent", then, relates to the latent learning, which occurred in preexposure phase but became overt in subsequent phases. The term "inhibition" relates to the decrement in learning in the test phase. Such data had been treated, earlier, as habituation and not as a learning phenomenon.
LI procedures have varied considerably. However, all LI studies (e.g., Zalstein-Orda & Lubow, 1995; Lubow, Rifkin & Alek, 1976; see a review by Lubow, 1989) use a preexposure phase (first phase) followed by a test phase (second phase). In the preexposure phase, a stimulus is presented with no consequence. In animals and in young children, at least according to Conditioned Attention Theory (CAT; Lubow, Weiner & Schnur, 1981; Lubow, 1989), this is a sufficient condition for decrementing the attentional response to that stimulus. However, in adult humans, a "masking task" usually is required in the preexposure phase in order to produce LI in the test (an issue that will be discussed later).
Basically, LI studies can be divided into three main categories:
1) Studies based on classical conditioning paradigms. These studies usually consist of three phases, preexposure, CS-US acquisition, and test, and LI is indexed by less disruption of behavior in the test phase by the stimulus preexposed group (PE) compared to the non-preexposed group (NPE). The difference is attributed to poorer learning by the PE compared to NPE group in the CS-US acquisition phase. This method has been used both with animals and humans. A series of conditioned taste aversion studies (e.g., in animals: Garcia & Koeling, 1967; Nachman, 1970; Revusky & Bedarf, 1967; in human studies: Arwas, Rolnick and Lubow, 1989) provides a typical example of an LI procedure with classical conditioning. Preexposure of a particular taste quality (CS; e.g. sacharine) interferes with the acquisition of food aversion (CR) when that same taste is paired with a toxin (US; e.g., LiCl) that produces a gastrointestinal disturbance (UR).
2) Studies based on instrumental conditioning paradigms. Such procedures usually consist of only two phases, preexposure and test, because the acquisition phase is actually combined with the test phase (during the association-acquisition phase, the behavior increases or declines). LI is characterized, in such a procedure, by slower conditioning. These methods are also used both with animals and humans. A study by Ackil and Mellgren (1968) provides a typical example of an LI procedure with instrumental conditioning. In preexposure, rats hear a tone with no associated consequence. In the subsequent phase, this tone is paired with shock. If the animal crosses from one side of the apparatus to the other side before shock onset, the shock is avoided. Rats preexposed to the tone acquire the avoidance response more slowly than rats not preexposed to the tone. (e.g. in human studies: De La Casa and Lubow, 1999).
3) Studies based on associative rule learning procedures. The third category, used only with human subjects, employs learning tasks which are not based on a conditioning paradigm. These are discrimination learning tasks used only with human subjects. The tasks can be solved by “trial and error” and the feedback given by the experimenter for correct and incorrect guesses. Here, there are only two phases, preexposure and test. LI is indexed by the number of trials required to solution, and the effect is demonstrated when the PE group requires more trials to reach the learning criterion than the NPE group. Zalstein-Orda and Lubow (1995) provide a typical example of such a procedure. In preexposure, the same meaningless shape (preexposed stimulus) appeared many times accompanied by a trigram. The masking task was to determine the number of times the list of trigrams repeated itself. In the test, the masking stimuli continued to be present on each trial, but on any given trial either the preexposed shape or a novel shape might appear. The subject had to learn that a change in the numerical value of a counter was associated with the presence of the previously irrelevant, preexposed stimulus. Subjects who experienced the preexposure phase with the to-be-relevant stimulus reached the learning criterion more slowly than subjects who experienced a novel stimulus in the test phase, thereby demonstrating the LI effect.
Various theories have been proposed to explain LI. Lubow (1989) has reviewed most of them, together with their weaknesses and strengths. (E.g., conditioned inhibition [Rescorla, 1971], competing responses [Lubow & Moore, 1959], Rescorla-Wagner model [Wagner & Rescorla, 1972] and others). Although there have been many explanations of LI, current theories fall into two major categories, association-deficit and retrieval-interference. The first is represented mainly by Conditioned Attention Theory and the second by Retrieval Failure Theory. Related to these theories are the two variables that have been extensively studied in regard to LI – context and masking.
Conditioned Attention Theory
Conditioned Attention Theory (CAT; Lubow, 1989; Lubow, Schnur, & Rifkin, 1976; Lubow, Weiner, & Schnur, 1981) states that "nonreinforced preexposure to a stimulus retards subsequent conditioning to that stimulus because during such preexposure the animal learns not to attend to it. The theory is based on the use of attention as a hypothetical construct, with the characteristics of a Pavlovian response, and on the specification of reinforcement conditions that modify attention" (Lubow, 1989). Attention is treated here as any other response that can be reinforced and conditioned.
There are several assumptions made by CAT concerning the attentional response (Lubow, 1989). With repeated stimulus presentations the attentional response declines. This decline is subject to conditioning. Attention can be modulated during the repeated presentation of the stimulus; it can be temporarily increased (by adding a novel stimulus with the target stimulus), an effect that also can be conditioned. Another assumption is that a minimal level of attention is required in order to create association to that stimulus. As attention increases, stimulus associability (the ability of that stimulus to acquire new associations) increases. The presentation of a nonreinforced stimulus will result in a decrement in attentional response to that stimulus. This decrement can be conditioned: the stimulus serves as CS, the nonreinforcement serves as the US which results in attention decline (inattention-UR). This decline is derived from the repeated stimulus presentations and obeys the laws of classical conditioning. CAT claims that increases and decreases of attention are on a single continuum. They are both responses resulting from the pairing of the CS with the presence or absence of reinforcement. Increases of attention result from excitatory US (e.g., shock) and decreases of attention from US absence (nonreinforcement). The organism responds to the US either by increasing or decreasing the attentional response.
CAT presents an innovative approach to attention. It makes a number of testable predictions which will be discussed in the next sections of the present study. The main predictions are: LI will be relatively long-lasting (a delay between preexposure and test still results in LI); LI will be relatively stimulus specific and will give rise to a stimulus generalization gradient; LI increases with an increase in the number of stimulus preexposures; LI increases as the inter-trial interval during preexposure increases; LI increases as the stimulus intensity during preexposure increases. LI can be disrupted if an external stimulus is presented in conjunction with the preexposed target stimulus. LI is subject to extinction, spontaneous recovery, blocking, and overshadowing.
CAT’s assumptions and predictions, which have integrated a variety of empirical findings, have also generated a substantial amount of research. Nevertheless, at least one issue remains problematic. According to CAT, attention reduction to the preexposed stimulus results in a relatively permanent interference with future associability of that stimulus. If that is the case, the effects of stimulus preexposure on the subsequent CS-US association should be represented in the test phase, and they should be relatively independent of the delay between acquisition and test. However, this is not the case. Several studies have reported a recovery of the CS-US association in test (see below).
Retrieval failure theory in LI
Retrieval failure theory claims that there is no impairment of stimulus associability, as claimed by CAT, but rather that preexposed and non-preexposed groups enter the CS-US acquisition phase with the same capability for forming new associations with the test stimulus. Instead, in the test phase, the association formed in the preexposure competes for expression with the association formed in the acquisition phase (Bouton, 1993; Miller, Kasprow & Schachtman, 1986).
Experiments by Miller, Kasprow and Schachtman (1986) have suggested that the relatively poor test phase performance (LI) is derived from an inability to retrieve the information encoded during the CS-US acquisition phase. This contradicts an important component of CAT because the latter claims that associability is reduced in the acquisition phase (as a result of unreinforced stimulus preexposure), and, therefore the CS-US association cannot be manifested in test as it was never acquired. Support for retrieval failure theory is derived from LI experiments in which the purported associative deficit was reversed; in the PE group, the CS-US association was manifested several days (12-20) after acquisition, although it was disrupted when the test phase immediately followed acquisition. In other words, LI was present with a short acquisition-test delay but not with a long delay; and this has been attributed to the recovery of the CS-US association during the long delay.
The basic claim of Miller et al. (1986) is that disrupted performance in test (LI) does not necessarily indicate the absence of prior CS-US association. We can never know for sure if learning occurred or not until it is manifested (an example of trying to prove the null hypothesis). The psychological literature is replete with examples of learning that, under a specific set of test conditions, is not overtly visible (e.g., defense mechanisms that generate suppression of knowledge). Cat’s claim that stimulus preexposure interferes with learning in the CS-US acquisition phase (second phase) is difficult to test directly. However, if performance in a test phase recovers, as for example with the passage of time between acquisition and test, this is evidence that acquisition did take place. Such recovered associations have been reported in several LI studies, thereby supporting, at least in part, retrieval failure theory and undermining the CAT explanation of LI.
Miller et al. (1986) claim that recovery of memory after forgetting is a consequence of either “spontaneous” recovery or a “reminder” treatment. They review a number of paradigm in which such effects take place, including retroactive interference, directed forgetting, extinction, infantile amnesia, retroactive clinical amnesia, paradoxical sleep deprivation, overshadowing, blocking, and others, including latent inhibition.
Although not many LI studies have investigated this issue directly, there is a considerable literature on retrieval and memory recovery. In regard to LI, Kraemer and Roberts (1984) observed spontaneous recovery from a CS-preexposure deficit after a 21-day retention interval in rats with a taste aversion task. These results were replicated by Kraemer et al. (1988) and Kraemer, Randall, and Carbary (1991). Kasprow et al. (1984) also reported attenuated LI of an auditory CS-footshock association in a fear conditioning procedure. They observed increased suppression when a reminder of footshock was presented outside of the test context. The reminder effect was not found in control groups lacking CS preexposure, or when the CS and US were presented in an unpaired relationship. Nevertheless, there are also a number of studies that have failed to produce a recovery effect. Although no variable has been identified which predicts recovery with certainty, Bouton (1993) and Miller et al. (1984) have concluded that LI is due, at least in part, to retrieval-failure. As mentioned before, this argument contradicts a major claim of CAT (the associability of the CS results in an impaired acquisition of CS-US association).
The role of context in LI
A theory of latent inhibition must take into account the role of context on the preexposure effect. First, LI is context-specific. That is, in order to produce LI, the test stimulus must appear in the same context as it did in preexposure. A change in context from preexposure to test disrupts LI (e.g., Bouton & Brooks, 1993; Channell & Hall, 1981, 1983; Honey & Good, 1993; Lovibond, Preston & Mackintosh, 1984; Lubow, Rifkin & Alek, 1976). Second, stimulus and context cues must be preexposed together in order to obtain LI. Third, context (alone) preexposure enhances LI, and context exposure (alone) which follows the preexposure of context-stimulus presentation does not affect LI. Although context effects were not part of CAT as first presented (Lubow, Schnur, & Rifkin, 1976; Lubow, Weiner, & Schnur, 1981), a review by Lubow, (1989) integrates the context-LI findings into CAT. In that review, it was proposed that context may serve as an occasion setter (Holland, 1985; Ross, 1983; Ross & Holland, 1981) for the stimulus no-consequence relationship, i.e., the presence of the same context as that of preexposure activates the stimulus- no-consequence association. If the context is a signal that retrieves the relation between CS and US, then when the context is changed, the activation of the association does not occur and LI is not evident.
Wagner (1976, 1978) claimed that stimulus presentation on the background of a familiar context "primes" the stimulus (CS) into short-term memory. As the to-be-conditioned stimulus is familiar, it receives reduced attention which, in turn, disrupts the rehearsal of CS-US association. As opposed to Wagner, Lubow and Gewirtz (1995) suggested that context serves to prevent the stimulus from entering into short- term memory by allowing for the expression of the stimulus-no-consequence association, thereby ensuring that the stimulus continues to be processed in the automatic mode. However, it was suggested that the occasion-setting role of context is to maintain the stimulus- no-consequence association, thereby preventing new associations from being developed to that stimulus (e.g., Lubow and Gewirtz, 1995).
Lubow and Gewirtz (1995) note that context has become a major issue in learning theory in general, and specifically in LI, both animal and human (Bouton, 1993; Lubow, 1989; Wagner 1978, 1981). For example, Zalstein-Orda and Lubow (1993) examined the effect of context in modulating LI in human subjects. The basic procedure consisted of a preexposure session in which the target stimulus (a nonsense shape), the context (the background of the computer screen), or both were presented under conditions of masking (determining the number of times a list of trigrams repeated itself). In the test phase, the subject had to learn that a change in the numerical value of a counter was associated with the presence of the previously irrelevant, preexposed stimulus. Change in context from preexposure to test abolished LI. It was also demonstrated, as in animal studies, that LI is dependent on conjoint preexposure of stimulus and context.
“Bouton (1993) claims that the context specificity of latent inhibition is consistent with a retrieval view. It is possible, for example, that an animal also treats the CS as familiar in a different context but that it fails to remember the significance of the stimulus (retrieval problem). Therefore, a reminder cue can retrieve the meaning of the association, the acquisition of which was never impaired.”
The role of masking in LI
Typically, in human LI studies, a masking task accompanies the preexposure of the to-be-target in the preexposure phase. The role of masking task in preexposure is to attract the subject's attention away from the target stimulus. Lubow (1989) summarized the major masking task issues: 1) There are some procedures (e.g., electrodermal conditioning, conditioned taste aversion) that may not require masking for the production of LI in adult humans. 2) Standard stimulus preexposure procedures (without masking) produce LI in young children (and in animals), but not in older children or adults. 3) If these same procedures are coupled with a masking task, then LI also can be produced in older children and adults.
Three reasons account for the necessity of masking (Lubow & Gewirtz, 1995). First, the presence of masking task stimuli in test modulates test task difficulty, preventing a floor effect (by making the test task more complex, differences in speed of solving the task can be observed). This, of course, also increases the chance for obtaining a ceiling effect (the subject completely fails to solve the task). Indeed, some of the failures to find LI in older children and adults with non-masked preexposure conditions were partly related to floor effects. Second, older children and adults treat the experimental environment differently than young children and animals. The former tends to allocate attention to any stimulus in the experiment apparatus, even if this stimulus is nominally irrelevant. They understand that the irrelevant stimulus might become relevant, and will allocate attention to it, unless a masking task attracts attention away from it. Young children and animals are less sophisticated concerning the experimental environment. They rapidly lose interest in the repeated irrelevant stimulus, and attention to it declines. It is clear that in order to encode the preexposed stimulus as irrelevant, attention must be diverted from it. In young children and animals, this appears to occur without any additions to the experimental situation. In older children and adults, the masking task is required to prevent the subject from allocating controlled attention to the stimulus. Third, masking reduces verbal mediating responses to the preexposure stimulus, which again may serve to maintain attention to that stimulus. While young children and animals react automatically to stimuli, older children and adults tend to mediate their response by verbal labeling, e.g. "here is this shape again". This by itself causes the stimulus to become relevant and may interfere with the development of the stimulus no-consequence association.
The role of attention in LI
It is clear that attention plays a critical role in LI (see CAT section above). The masking task requirement in LI studies with adult humans, as described, attests to that role. At the same time, the LI procedure affects attention. The preexposure manipulation produces a decrement in the attentional response to the preexposed stimulus (at least according to CAT), thus limiting the associability of that stimulus in the test phase. Indeed, Lubow and Gewirtz (1995) claim that LI has often been construed as reflecting the operation of attentional mechanisms, and that "LI has recently become the linchpin of certain animal models of human attentional dysfunctions". CAT’s fundamental assumptions, as described, are that attention and conditioning processes govern LI. To repeat, during preexposure, an attentional response is first elicited by the novel target stimulus, and then it gradually declines with repeated presentation of that stimulus, resulting in the development of an inattentional response. However, it is important to note that most LI studies do not employ standardized, controlled attentional manipulations. One exception is a recent study by Braunstein-Bercovitz and Lubow (1998) in which they tried to determine whether the repeated stimulus presentation of the LI preexposure procedure uses any attentional resources, or whether it is an attention-free process. In other words, is LI governed by controlled or automatic processes. Shiffrin and Schneider (1977) defined automatic processes as fast, parallel, effortless, and not directly controlled by the subject. Controlled processes are the opposite. They are effortful, capacity limited, operate slowly and serially. Another distinction between automatic and control processes is that the latter is a voluntary “top-down” process which enables the individual to direct attention to a particular location without a visible cue at that location. Automatic processes, on the other hand, are “bottom-up”, in which attention is allocated automatically to target detection (Krose and Julesz, 1990). Braunstein-Bercovitz (1997) manipulated attention by varying masking task load. It was assumed that as more attentional resources are devoted to the masking task, fewer resources would be devoted to the target stimulus. As expected, LI was affected by the load manipulation. First, it was demonstrated that a masking task was necessary to produce LI. Second, LI was produced only when sufficient attentional resources were available for the processing of the preexposed stimulus. Third, LI was produced only when repeated presentations of the preexposed stimulus resulted in a decline of the attentional response to it (manipulated by the location of the stimulus, central or peripheral to the masking task letters).
It can be concluded, then, that in order to obtain LI, the preexposed stimulus (which becomes the target in test) should receive some minimal amount of attention, but that this attention should decline with time passing and become automatic rather than controlled.
In spite of the important role of attention in LI, the customary LI procedure does not allow for a direct measure of attention. The traditional test-phase procedure uses a learning score such as number of correct responses or number of trials to a learning criterion. These measures have a number of problems, including the fact that they are discrete and that they often generate floor or ceiling effects. To assess attention more directly, a new approach was needed.
Context-target relation in LI
Lubow, Rifkin and Alek (1976) used a procedure similar to that of Zalstein-Orda and Lubow (1995), but they manipulated relation between a "target stimulus" and "environment stimuli" (also referred to as context, distractor, or background stimuli). The target stimulus was the typical preexposed stimulus, as mentioned before. In this case, it was either a set of round or angular objects placed in front of the subject. The environmental stimuli were all of the stable, unchanging stimuli surrounding the subject (the shape and color of the table, chairs, and walls). For short, we will call the environments red and blue, according to which of the two environment colors the subjects were exposed to. Subjects received a specific combination of environmental and target stimuli, red-round, red-angular, blue-round, or blue-angular, which they observed with no consequences. For the test phase, each of the four combinations generated four conditions of familiarity: same environmental and target stimuli as in preexposure (old stimulus and old environment = SoEo); same environmental stimuli but a novel target stimulus (SnEo); novel environmental stimuli and the same target stimulus (SoEn) and novel stimuli both in environment and target stimuli (SnEn). The procedure did not include a masking task, as the subjects were small children in one experiment and rats in another experiment. The LI effect was demonstrated by a retardation of learning in subjects tested in the SoEo condition compared to subjects in the SnEo condition. However, the four conditions produced additional findings related to other paradigms. For example, perceptual learning (e.g., Gibson & Walk, 1956; Epstein, 1967; Oswalt, 1972) was demonstrated by better performance in the SoEn condition as compared to the SnEn condition. Importantly, this set of findings solved a major paradox in the learning literature, namely that in some instances, preexposure resulted in improved performance (perceptual learning) and in other instances preexposure resulted in retardation of performance (LI). By separating contextual and focal stimuli, they were able to conclude that "the most effective conditions for learning are those in which there is a contrast between the novelty of the environment and the novelty of the stimulus", irrespective of whether it is a new stimulus in an old environment or an old stimulus in a new environment. Kaplan (1992) found additional support for these conclusions.
The present study
Introduction: Considerations regarding context-target relationships which are necessary for LI (Lubow, Rifkin & Alek, 1976), the need to build a within-subject design for measuring LI, and the need to develop a procedure that indexes the attentional aspect of LI more directly, all pointed in a direction which produced the “the visual search approach to the study of LI”. The new procedure consists of two phases, preexposure and test. In both phases, the subject takes part in a visual search task in which he has to detect the presence or absence of a unique target among homogenous distractors. In the preexposure phase, all of the trials are homogenous (across trials, the distractors-target presentations are the same). In the test phase, the trials are heterogeneous (each trial may consist of different/same distractors and/or different/same target compared to preexposure).
The context-target relations: The concept of context-target relation is the same here as in Lubow, Rifkin and Alek (1976), but instead of using context and target stimuli which are not of the same type (in Lubow, Rifkin and Alek, context was primarily environment color, and target was mainly object shape), the contexts and targets were qualitatively similar. Context stimuli were identical nonsense shapes (distractors). The target also was a nonsense shape, but different than the distractor. This context-target relation allows one to examine what happens when the context stimulus (nonsense shape) is converted into a target stimulus in a later phase, and vice versa. The preexposure phase in the present visual search procedure is similar to other LI procedures, especially to that of Lubow, Rifkin and Alek, (1976). The subject performs a task in the presence of distractors and a target. As noted, the present design allows one to investigate the effects of distractor and stimulus status during preexposure on test performance. Thus, the target stimulus in test may be congruent with its status during preexposure, a target in both phases: T->T. It can also be incongruent, a distractor during preexposure becoming a target in test: D->T. In addition, the test target may be also novel: N. Similarly, the test distractors may be congruent: D->D, incongruent: T->D, or novel: N. This design creates nine (3 x 3) possible distractors-target combinations, only seven of which can be realized. T->T:T->D and D->T:D->D (in these and the following expressions, the left hand side of the colon indicates the status of the target at the time of test; the right hand side of the colon indicates the status of the distractors at the time of test) are the two combinations which cannot be generated since they result in presentations in which distractors and target are identical and therefore a unique target cannot be displayed.
In general, this design allows one to create a variety of familiarity conditions, some of which have never been examined before. Of particular importance are the conditions that replicate the traditional LI procedure. From earlier descriptions of such procedures which employ masking (e.g., Zalstein-Orda and Lubow, 1995), it is apparent that during preexposure the masking task stimuli are equivalent to targets and the to-be-associated but presently irrelevant stimuli function as distractors. In the test, the masking stimuli, which are still present, become the distractors, and the preexposed irrelevant stimuli become the target. Thus, the stimulus preexposure group can be represented as D->T:T->D. The group that is not stimulus preexposed also receives the masking task during preexposure and test, but the test target is novel, thereby creating the N:T->D condition. Therefore, performance in the D->T:T->D compared to N:T->D condition is equivalent to the typical comparison of PE versus NPE (preexposed versus non-preexposed) groups in traditional LI studies.
Another comparison of interest is the N:T->D and N:N condition, which represent conditions for demonstrating a novel pop-out effect (Johnston, Hawley, Plewe, Elliott & DeWit, 1990). Novel pop-out (NPO) occurs when a new target which is presented in an array of familiar figures (N:T->D) is detected faster than a new target in an all-novel array (N:N). This effect is of particular interest because LI and NPO share a common condition, N:T->D. This condition, which is responsible for eliciting NPO, corresponds to that of the non-preexposed group in the LI paradigm. In other words, the control or baseline condition in LI is the experimental one in NPO. This suggests that LI effects may arise from two independent sources. In addition to the stimulus preexposure group generating poor performance as a result of decreased attention to that stimulus at the time of test, the non-preexposed group may induce superior performance, because at the time of test the target is a novel stimulus presented in a context of familiar distractors.
In summary, the new procedure and design enable one to generate several familiarity conditions in test, some analogous to traditional LI procedure, and some to conditions for eliciting NPO, as well as several others. For the purpose of the present study the three conditions of special interest are D->T:T->D, N:T->D, and N:N, where the first and second conditions represent LI, and the second and third, NPO.
The within-subject design issue: As already described, most LI studies reported in the literature are based on between-subject designs, i.e., the preexposed condition and the non-preexposed condition are represented by independent groups. The nature of the test task is such that if the subject participated in one condition he/she can no longer be a subject in a different LI condition (e.g., see test phase in Zalstein-Orda & Lubow, 1995; Lubow, Rifkin & Alek, 1976).
The inability to compare the same subject in different conditions (within-subject design) can be very costly, especially in cases where the experiment utilizes with special populations, a common situation with LI studies of psychopathological groups, such as schizophrenics, ADHD children, Parkinson patients, etc. (e.g., Salzman, Hadar, Korczyn, and Lubow, 1984; Lubow and Josman, 1993; Baruch et al., 1988; Gray, Hemsley and Gray, 1992). From a statistical point of view, a within-subject design is also more efficient as it reduces the variance between the experimental conditions.
The visual search paradigm allows one to use a within-subject design because the test task does not involve rule learning. Rather, on each trial the subject has to detect whether the target is present or absent among the distractors. The detection of the target on one trial does not prevent the subject from continuing the task on the next trial. The correct response on one trial does not predict the correct response on the next trial, as it does in LI learning tasks. As a result, the subject can be exposed to various types of trials in the test, including the two conditions of interest for LI, D->T:T->D, N:T->D (as well as NPO). Performance on D->T:T->D trials and N:T->D trials can be compared in a within-subject design, thereby solving many of the aforementioned problems inherent in typical LI studies.
The nature of the visual search procedure: As mentioned, CAT, which is the main theory for explaining LI, emphasizes the role of attention in LI. Nevertheless, most LI studies define the dependent variable as number of trials to a learning criterion, a measurement which, at best, is only an indirect measure of attention. One of our goals was to use a task that measures attention more directly. Such a measure can be obtained from visual search response time. All of the studies to follow will employ such a task, both in the preexposure and in the test phases. The visual search analog of LI provides additional advantages over the older procedures, as described below.
Posner and Boies (1971) indicate that the study of human attention may be divided into three components, alertness, selectivity, and processing capacity. Alertness is treated as the ability to develop and maintain an optimal sensitivity to external stimulation. It is often mentioned in the context of human ability to perform on long, boring tasks like those that psychologists design to study vigilance (Machworth, 1970). Selectivity is the ability to select information from one source rather than another (Egeth, 1967; Triesman, 1969). Processing capacity relates to the idea of a limited central processing capacity (Broadbent, 1958).
Visual search tasks provide conditions to assess these three components of attention. In these tasks, the observer has to report the presence of a predefined target pattern in a display consisting of many background elements. Target detectability depends on properties of the target and background elements and the positional uncertainty of the target (Krose and Julesz, 1990). Accuracy and speed of detection are the main indicators used to assess performance in visual search tasks. The quality of performance is dependent on the three elements mentioned: A) level of alertness – depends on the particular state of the observer and the nature of the task (how long, boring etc. the task is). B) Selectivity – depends on the particular ability of the observer to handle visual-spatial information and on the structure of the task. For example, Krose and Julesz (1990) claim that if, on the one hand, the target and the background elements are very dissimilar, then detection is effortless – the target seems to “pop-out”. If, on the other hand, they are similar, then detection is more difficult. C) Processing capacity – Depends mainly on the complexity of the task and its demands in regard to dividing attention.
easy. A RT ceiling effect can be prevented by using a self-terminated-response task, so that the subject who finds a condition more difficult has unlimited time to respond.
Demand constancy: The similar demands in preexposure and test (visual search in both phases) maximizes the context constancy across the two phases of the procedure. This was not present in previous LI studies where preexposure and test had different task demands. Such constancy is important, as LI is context-specific. As mentioned before, in order to obtain LI the stimulus must appear in the same context in preexposure and test. Since task demands are part of task context, the proposed procedure should potentiate LI effects.
Floor and ceiling effects: Some LI studies have suffered from problems of floor or ceiling effects in test. Such effects occur when the test learning task is either too difficult or too easy. The ceiling effect is especially problematic for populations in which a “super-LI” effect is expected (e.g., Parkinson disease patients in the study of Salzman, Hadar, Korczyn and Lubow, 1994). The floor effect might be as problematic as the ceiling effect, but here the problem would derive from the task being too easy, preventing any learning differences between preexposed and non-preexposed groups.
In the visual search procedure such floor and ceiling effects are not problems. First, the nature of the procedure is such that the subject does not benefit from a correct response; the problem from engaging in a learning task does not exist. On the other hand, there is no floor effect with RTs, if sufficient distractors appear in the presentation so that the task is not tooObjectives*
The experimental portion of the dissertation consists of six studies, all based on the same basic visual search paradigm. Experiments 1 and 2 concentrate on the development of the visual search procedure which represents a new approach to LI. The first experiment aimed at actualizing and validating the new approach, and producing the LI effect. The main hypothesis was that RT in the D-sensitive to context change as in traditional LI studies?
The contradictory results of Experiments 1 and 2 raised central theoretical issues for the visual search approach to the study LI, as well as LI in general. As will be described later, these results have important implications for major controversies in the LI literature. Therefore, in the second experimental section (Experiments 3-6), we focused on the following theoretical issues: the debate between conditioned attention theory and retrieval failure theory; the effects of reminder-trials on LI; the effects of context on LI; the relation between LI and novel popout.
Experiments 3, 4, and 5 attempt to optimize the LI-visual search procedure while at the same time trying to resolve the aforementioned theoretical issues. The first hypothesis in these experiments was identical to that of Experiments 1 and 2, namely that there would be an LI effect. An additional hypothesis was that the “reminder” condition is critical for the actualization of the LI preexposure effect (i.e., the presence of reminder-trials will result in LI, while the absence of reminder-trials will attenuate LI). The results of Experiment 3 and 4 suggested that reminder-trials indeed, are critical for the actualization of the LI preexposure effect, but only when there are heterogeneous test trial conditions.
Experiment 5 used a procedure to ensure that the LI found in the previous experiments was derived from preexposure effects and not from within-trial test conditions, such as negative priming. The hypothesis was that the effects found in test would be abolished when the preexposure phase is omitted, or with a “dummy” preexposure phase (one in which neither the preexposed target nor the distractors appear on any of the test trials).
Experiment 6 returned to the “starting point” (Lubow, Rifkin and Alek, 1976; Zalstein-Orda and Lubow, 1995) and investigated the effects of context change on LI in the new visual search paradigm. It was hypothesized that color context change would reduce the LI effect, as expected from results of traditional LI studies.
Each experiment is followed by a discussion section and an introduction to the next section. A general discussion follows the last of the six experiments.
Results section: omitted
Summary of goals and results
The main purpose of the present study was to develop a new approach to LI, based on a within-subject design, and on a visual search procedure which measures attention more directly than traditional LI procedures. During the development of the procedure, a number of critical theoretical questions concerning the LI mechanism in visual search were raised:
Six experiments, all using a two-stage visual search procedure, were conducted. In addition to evaluating the effects of stimulus preexposure on test phase responding, effects of reminder and context manipulations on the preexposure effect were examined.
Experiment 1 evaluated the basic experimental procedure. The preexposure phase used presentations of homogenous distractors and target. This was followed by the test phase which contained heterogeneous presentations, consisting of seven different familiarity conditions. Familiarity level was defined as target being familiar as a result of its former preexposure either as a target, or a distractor, or not familiar (novel). In the same manner, the distractors could be familiar as a result of either preexposure as a target, or a distractors, or not familiar (novel). An LI-like effect, defined as a difference in RT between PE and NPE conditions, was demonstrated. In addition, an NPO effect, defined as the difference in RTs between NOV and NPE, was obtained. Additional analyses were performed in order to identify the type of processing used in the visual search task. RTs on non-target trials were longer than on target trials, both in preexposure and test. In addition, it was demonstrated than RTs in the preexposure phase were shorter than RTs in test phase, and that learning curves in the preexposure phase were steeper than in the test phase. For the non-target conditions in test, the learning curve was flat.
In Experiment 2, the seven familiarity conditions were reduced to four. The target in test could either be familiar as a result of being a distractor in preexposure, or it could be novel. The distractor in test could either be familiar as a result of being a target in preexposure, or novel. Surprisingly, the robust LI effect found in Experiment 1 was absent, except on the first test trials. It was suggested that the discrepancy between the two experiments may have resulted from omitting one of the Experiment 1 test conditions from Experiment 2, specifically the condition that could serve as a reminder cue (in which target and distractors appear identically in test as in preexposure). A review of the literature indicates that such a reminder treatment may be important for producing LI. The reminder cue may serve both as a cue for retrieving the stimulus-no consequence association which was formed in the preexposure and/or it may help to maintain context constancy from preexposure to test. Both possibilities, which may be related to each other, have received some support, particularly in the animal-LI literature.
Experiment 3 was designed to cope with the conflicting results of Experiments 1 and 2. Specifically, the reminder condition was manipulated so that in the test phase it either did not appear (0-reminder) or it appeared on 25, 50, or 75 percent of the test trials. To enable a between-subject analysis of the results (in addition to a within-subject analysis), each test condition was presented in a separate block. In this way, comparing across blocks provided the basis for a within-subject analysis. However, if one isolates the first block of each subject, then between-subject comparisons can be made. With this procedure, an LI effect was found, as it was in Experiment 1. Accordingly, the analyses of Experiment 3 were conducted twice, once with data from the within-subject design, and once with data from the between-subject design. Both analyses provided a similar pattern of results, thus supporting the proposal that the within-subject visual search design taps the same process as traditional between subject LI designs.
LI-magnitude was a function of percent of reminder trials (0, 25%, 50%, and 75%). A strong LI effect was found, as expected, in the condition with the largest percent of reminder trials (75%). However, the most striking result was that the non-reminder group, for which it was predicted that LI would be abolished, demonstrated a significant LI effect, even larger than in some of the reminder groups.
The unexpected results may be a result of differences in test task structure. Specifically, test task in the reminder groups may have produced a higher level of task difficulty than the test task in the non-reminder group. Within each block, reminder group subjects were presented with two conditions, the reminder trial plus the trials from one additional condition (either PE, NPE, or NOV). On the other hand, the no-reminder group had only one test condition (either PE, NPE, or NOV). In Experiments 1 and 2, the test task was also difficult as the non-blocked procedure generated many familiarity conditions (seven and four, respectively). In those situations, the reminder treatment was deemed necessary to retrieve the association that was acquired in the preexposure phase. Thus, the reminder cues in Experiment 1 generated LI and its absence in Experiment 2 precluded LI. However, in Experiment 3, the test conditions were presented in isolated blocks, thus providing fewer sources of between-trial interference than those of previous procedures. It seems that in such circumstances the reminder cue may change the simple test task into a more difficult one, and thereby attenuate LI. This change in task difficulty may have been the source of the unexpected presence of an LI effect in the no-reminder group.
Experiment 4 was designed to integrate the results of the three previous experiments. It consisted of a procedure which, on the one hand, resembled those of Experiments 1 and 2 – the test task was again heterogeneous, but, as in Experiment 3, it also manipulated the reminder trials. The subjects in the reminder group received four test conditions, reminder, PE, NPE, and NOV. The subjects in the no-reminder group received the same four conditions, but instead of the “real” reminder, they were presented with a novel condition in which the targets and distractors were not present in the preexposure phase. As predicted, the LI effect was stronger in the reminder group than in the non-reminder group; this difference increased across blocks.
Experiment 5 was essentially run as a control for the basic preexposure effect. The subjects were treated exactly as in the reminder group of Experiment 4, only the preexposure phase was composed of “dummy” trials. Target and distractor stimuli which appeared in preexposure never reappeared in test. As the meaning of PE, NPE and NOV is defined according to the relationship between preexposure and test displays, the dummy preexposure manipulation results in all test conditions being novel. As the order of the test conditions was fully counterbalanced, no effect was expected, nor was any found. These results confirm that the effects found in previous procedures were, indeed, due to the preexposure procedure which generated the PE, NPE, and NOV test conditions.
After achieving the main purposes of the dissertation – designing a visual search, within-subject procedure that generates a LI-like effect, and identifying the role of reminder cues in test, the last Experiment sought to integrate the visual search approach with manipulations which have been found to be effective in traditional LI studies. The idea for varying distractors and target familiarity was derived from Lubow, Rifkin, and Alek (1976) where LI was a function of the novelty/familiarity relationship between context and target. In the present study, distractors were equivalent to the context stimuli of Lubow, Rifkin, and Alek (1976), with the advantage that the distractors could be drawn from the same class of stimuli as the targets (nonsense shape). This was not possible in the Lubow et al. study, as context stimuli were totally different from target stimuli (environmental color versus object shape). On the basis of studies by Zalstein-Orda and Lubow (1995) and Kaplan (1992), it was suggested that both context and distractors-target manipulations can be integrated into the same procedure, i.e., that context in the conventional meaning of stable, unchanging stimuli of the situation, can be manipulated in addition to the manipulation of the discrete stimuli which generate the visual search procedure. Experiment 6 did exactly that. The basic procedure was the same as that of the Experiment 4 reminder group. In addition, context color was manipulated. Subjects were exposed to either a red or blue screen color while performing the visual search task in preexposure. In test, the screen color either remained the same as in preexposure or was changed either from red to blue or blue to red. The constant context color (across preexposure and test phases) produced a stronger LI effect than the changed context color. The source of the difference was derived from the PE condition and not from the NPE condition.
What was actually learned in preexposure – The role of attention
As mentioned, attention plays a critical role in LI. CAT’s fundamental assumptions are that attention and conditioning processes govern LI. The preexposure manipulation produces a decrement in the attentional response to the preexposed stimulus thus limiting the associability of that stimulus in the test phase. During preexposure, an attentional response is first elicited by the novel target stimulus, and then it gradually declines with repeated presentation of that stimulus, resulting in the development of an inattentional response. LI is produced only when sufficient attentional resources was available for the processing of the preexposed stimulus and only when repeated presentations of the preexposed stimulus result in a decline of the attentional response to it.
In spite of the important role of attention in LI, the customary LI procedure does not allow for a direct measure of it. The visual search approach to the study of LI overcame this problem by using measurements of RTs in a visual search task, which is the most frequent method for measuring attention in human psychological research. This method is a more sensitive than “number of trials to criterion” which has been the default measurement in human LI studies until now. In addition, visual search has fewer methodological problems, such as ceiling and floor effects which have often plagued human LI studies.
In the visual search task the subject’s task was to detect a target among distractors. For performance to improve over trials, the subject had to learn to ignore the distractors. In this sense, in preexposure, the target stimulus is equivalent to the masking task stimuli used in classical LI studies. In both cases, the subject's attention is diverted from the to-be-tested stimulus (the preexposure distractors that appear later as a target in the PE test condition). One may ask, if it is really possible for the subject to ignore the distractors which are relatively salient in the visual search task? There are two related answers. Rock and Gutman (1981), for example, described a procedure in which subjects had to selectively attend to one of two overlapping shapes. The attended shape, of course, was later recognized as having been exposed previously. However, the subject was unable to recognize the unattended one. Therefore, the presence of the stimulus in the focal visual field did not necessarily result in the allocation of attention to that stimulus. However, as we know from the traditional LI studies, the nominally unattended stimulus does receive some processing. In order to learn to ignore a stimulus, the subject first must encode its properties, and only then can he develop stimulus-specific LI. This process was described by Lubow (1989), with a schematic representation of CAT, as part of an information processing system. He claimed that stimulus encoding precedes in time the conditioning of inattention. As noted, in order to learn not to attend to a particular stimulus, the characteristics of the stimulus must first be encoded and stored; the engagement of the process that leads to the stimulus-specific conditioning of inattention can only occur after initiation of the process for encoding stimulus properties. The absence of such encoding would result in an absence of the internal representation of the to-be-conditioned stimulus, and, consequently no LI (e.g., Braunstein-Bercovitz, 1997).
In the Lubow, Rifkin, and Alek (1976) study, in which context and target stimuli were qualitatively different, the processes responsible for LI and perceptual learning were separated. In that study, the differences between context and target were obvious (room color versus object shape). In the present study, distractors (context) and targets were quite similar. Therefore, it was expected that, at least at the beginning of the preexposure phase, attention would be allocated to both distractors and target stimuli, and that both would be encoded into long-term memory. It is only with continuing practice that the subject would learn not to attend to the distractors. This brings us back to Lubow’s (1989) claim that stimulus property encoding precedes the conditioning of inattention. Although this notion described processing according to CAT, it also may be relevant to the major premise of retrieval failure theory, that LI modulation is best understood by processes which occur in test. From this point of view, the subject reaches the test phase with two different sources of encoded information about the unattended stimulus, which compete with each other. One source of encoded information is derived from the early stages of preexposure, where the subject still allocated controlled attention to the to-be-tested stimulus. The other source is derived from the process, later in the preexposure phase, that encodes that to-be-target stimulus is not relevant. It is suggested, therefore, when after the preexposure phase terminates, and the test phase begin, the subject may deploy two alternative processing pathways. If the test conditions are such that the subject perceives that the "rules" did not change from preexposure to test, then the encoded information concerning the irrelevance of the preexposed stimulus will be activated and will generate the LI response pattern. Components that should enhance such perceptions include context constancy and reminder cues. On the other hand, the subject can perceive that the "rules" have been changed from preexposure to test, that the encoded information about irrelevance of the preexposed stimulus is no longer valid. In that case, the preexposed stimulus, whose stimulus properties already have been encoded, may even have an advantage over a novel stimulus for further processing. The result is better test performance with this stimulus, resulting in, for example, perceptual learning (e.g., Gibson & Walk, 1956; Oswalt, 1972; Epstein, 1967; Gibson, 1969), the mere exposure effect (Zajonc, 1968), and others, or at least, not resulting in LI.
Theories that refer to developmental changes in perception, e.g. Hochberg, 1981, 1982, suggest that during the developmental process, mental models (schemata) are constructed of the perceptual situation. The models consist of hypotheses about the external world. They operate so as to select relevant information, and then compare that information to existing schemata (Coren and Ward, 1989). Other theories use the term “schema”, a term which refers to a cognitive structure stored in memory which contains abstract representations of events, objects, and relationships in the external world. One can treat the preexposure experience in LI as contributing to the schema construction process. The subject acquires a schema of the preexposure situation together with retrieval cues for that schema. In the test phase, if contextual cues from preexposure, such as background color of the preexposure displays or reminders of preexposure phase trials (T->T:D->D), are available, the schema that was acquired during preexposure is retrieved and activated: the preexposed target (now the distractor) is relevant and should be detected, the preexposed distractors (now relevant) are irrelevant and should be ignored. This will result in poor performance since the demands of the test presentation are exactly opposite to those of the test (D->T:T->D). However, if the context cues from preexposure are changed in test, the subject has more difficulty in retrieving the preexposure schema. In that case, one obtains results such as in Experiment 2, in the no-reminder group in Experiment 4, and in the context color change groups in Experiment 6. In all of the above cases, LI was abolished or weakened.
The reminder trial as a cue for schema retrieval
The relevance of the reminder condition for eliciting LI is central to the present study. The experiments which involved reminder treatments produced different patterns of results in different experiments. When the test task was complex (Experiments 1, 2, and 4; several different test conditions in a random order), the reminder was necessary to produce LI. However, when the test task was simple (Experiment 3 – three-block design; each test block consisted of an homogenous condition), the reminder was not necessary, and it was even an obstacle to obtaining LI. In Experiment 3 LI was obtained in the absence of reminder cues. The current analysis suggests that when the subject is confronted with a complex test task which is different from the homogenous preexposure phase task, the reminder cue retrieves the preexposure schema. That schema defines which stimuli are relevant and which are irrelevant according to the internal representations encoded during preexposure. However, when a subject is confronted with a test which is homogenous and similar to preexposure, as in Experiment 3, the schema is easily accessible without additional reminder cues. In the three-block design of Experiment 3, the block structure is identical to that in preexposure (one homogenous condition). Here, the introduction of a reminder condition might even be detrimental to performance, as it adds to the difference between the test and preexposure conditions. In summary, there are four frameworks, in each one of which the reminder has a different role:
In the last several years much effort has been expended by LI researchers in designing a LI within-subject procedure. These efforts have not been successful. The discussion in the previous paragraphs may provide an explanation for these failures. Traditional LI studies have used between-subject designs in which the preexposure and test phases were composed of completely different tasks. In preexposure, the subject focused attention on the masking task, and in test on the stimulus preexposed in conjunction with a different task. This procedure does not contain a reminder/retrieval cue because the test contains either an homogenous D->T:T->D (PE) or an homogenous D->T:N (NPE) condition. In the between-subject design, this is a sufficient condition to elicit LI as described in the present study. We have seen that the no-reminder group in Experiment 3 produced robust LI, as was always the case in the traditional between-subject designs. However, when the two test conditions, D->T:T->D (PE) and D->T:N (NPE), are presented to the same subject in a within-subject design, LI is attenuated or even not obtained at all. It is suggested that the failure to obtain LI in within-subject LI design, results from the subject being in the situation described in paragraph 4 – “a mixed conditions task with no reminder cues”. The change of context from preexposure to test interacts with the new task complexity (difficulty), both of which interfere with the retrieval of the preexposure schema, thereby precluding LI. On this basis, it is suggested, that researchers developing within-subject designs in the traditional LI framework should use procedures that are similar in preexposure and test and that other preexposure reminder cues be introduced in the test.
CAT versus Retrieval Failure Theory: Learning, forgetting, and remembering.
Recovery of learning (forgetting preexposure – attenuating LI) and Conserving LI (remembering preexposure by reminder treatments)
The reminder manipulation effects emphasize the role of memory in obtaining LI, a factor that has not received sufficient attention in the LI literature. This observation is congruent with Bouton’s (1994) claim that “theories of learning and memory have been artificially separated for years. This separation is a result of historical reasons. When learning was viewed as the acquisition of an association between a stimulus and a response, any notion of memory beyond the stimulus-response association itself seemed unnecessary to explain behavior”. However, with the development of sophisticated cognitive tasks (e.g., rule learning), specifically those used with LI, there appears to be a need to introduce concepts drawn from the memory literature. Even Pavlov (1927) recognized that some learning effects are related to cognitive processes that cannot be explained by stimulus-response models. For example, he described effects such as “spontaneous recovery”, in which the extinguished response recovers if time elapses before, again, assessing the response to the CS. The fact that the original response can be recovered after extinction indicates that the initial learning could not have been eradicated, but instead was lying dormant in memory.
Although LI and extinction may not be governed by equivalent processes, they share a common component – a decline in performance. More importantly, they both exhibit spontaneous recovery, a fact which raises theoretical debates among researchers working with both paradigms. There are several LI studies that have demonstrated spontaneous recovery of learning (LI was abolished and learning was recovered and manifested in test), e.g., De La Casa & Lubow, 1995; Bakner, Strohen, Nordeen & Riccio, 1991; Kraemer, Randall, & Carbary, 1991; Kraemer & Spear, 1992; Kraemer & Roberts, 1984; Kraemer & Ossenkopp, 1986. There are also many studies on spontaneous recovery and renewal effects in extinction of conditioning, e.g., Pavlov, 1927; Brooks & Bouton, 1993; Rescorla & Cunningham, 1978; Robbins, 1990; Thomas & Sherman, 1986.
A better understanding of LI might be gained from examining that literature on recovery. The major variable that is used to explain recovery from extinction is that of context change. According to Bouton (1994), renewal and spontaneous recovery are due to a failure to retrieve the memory of the extinction episode. In both cases, the subject fails to recall the extinction episode outside the extinction context. If this is true, then presenting a cue that retrieves the extinction episode (reminder cue) at the time of test should reduce spontaneous recovery. Brooks and Bouton (1993) showed that spontaneous recovery indeed, can be attenuated by presentation of an extinction cue that had preceded some of the to-be extinction trials. Therefore, we can conclude that associative-stability is dependent on context. In all of the above cases, change of context from preexposure to test resulted in a retrieval failure, which could be reversed by a reminder treatment. As mentioned earlier, forgetting and remembering of encoded information is adaptive in that they decreases information overload from previously irrelevant stimuli, and enable the use of it when it again becomes relevant. The availability of such processes suggests that the preexposed information, whether relevant or irrelevant, is present, stored, and not lost. This brings us to the debate between CAT and retrieval failure theory. The reminder effect, demonstrated in the present study, support the position of the latter one.
As mentioned, CAT (Lubow, 1989; Lubow, Schnur, & Rifkin, 1976; Lubow, Weiner, & Schnur, 1981) states that "nonreinforced preexposure to a stimulus retards subsequent conditioning to that stimulus because during such preexposure the animal learns not to attend to it”. According to CAT attention reduction to the preexposed stimulus results in a relatively permanent interference with future associability of that stimulus. As opposed to CAT, Retrieval Failure Theory claims that in the test phase, the association formed during preexposure competes for expression with the association formed in the acquisition phase (Bouton, 1993; Miller, Kasprow & Schachtman, 1986). Experiments by Miller, Kasprow, and Schachtman (1986) have suggested that the relatively poor test phase performance (LI) is derived from an inability to retrieve the information encoded during the acquisition phase (of a three stage procedure – preexposure, acquisition, test).
As already mentioned, the results of the present study support the position of Retrieval Failure Theory that preexposure does not have a permanent effect on associative performance. Change of test conditions from preexposure conditions interact with the effects of preexposure per se. Task complexity and/or the absence of reminder cues in test may abolish LI and recover learning from preexposure association.
Thoughts about integrating CAT and Retrieval Failure Theory
As claimed earlier, the context in test is functioning as an occasion setter or as a reminder cue for the preexposure phase. According to the Retrieval Failure Theory (RFT), the CS-US association is acquired but not manifested in test because of retrieval failure. This failure persists as long as the context is constant, but can be recovered if the context is changed from preexposure to test phase. This argument is supported by the mentioned studies in which a context change in test resulted in retrieval of the CS-US association that was inhibited (LI). The recovery of the CS-US association in test contradicts CAT, because the latter claims this association was never acquired.
Although our data support Retrieval Failure Theory and not CAT, it may be possible to integrate the two apparently conflicting explanations of LI. To begin with, it is important to note that Retrieval Failure Theory is not specific to LI. CAT, on the other hand, was proposed to account only for the LI phenomenon. As a result, CAT has many components which are not related to the controversy and therefore can still be maintained even if certain claims of Retrieval Failure are correct.
At the core of the debate, we can differentiate between two arguments. The first is the claim of Retrieval Failure Theory that context change from preexposure to test can produce recovery from LI. This does not contradict CAT, and even has been integrated into it (Lubow, 1989). Both sides agree that context change abolishes LI. Furthermore, since the passage of time promotes change of context (the internal milieu of the subject changes, new experiences are encountered, new schemas are constructed, etc.), spontaneous recovery from LI after long delays between acquisition and test also can be treated as context change.
The second component of the debate is more problematic. CAT states that preexposure results in a decline in associability of the target stimulus, a deficit that should disable the acquisition of the CS-US association, the result of which is that in the test phase there is nothing to retrieve from the acquisition phase. Clearly, in an LI experiment, obtaining spontaneous recovery of the CS-US association would falsify the CAT position.
However, there may be several parallel processes that are normally operative during acquisition regarding the CS-US association. If this is the case, the manifestation of learning in test can be both reduced (CAT) and recovered (Retrieval Failure Theory). It might be useful to explain this notion by an example. It is well known that in cases of anterograde amnesia there is an inability to learn new information. Nevertheless, the well-known case of patient H.M who suffered from anterograde amnesia (Milner, Corkin, and Teuber, 1968; Milner, 1970; Corkin, Sullivan, Twitchell and Grove, 1981) demonstrated that although new overt associations could not be generated, some associative learning can occur. As an example, Cohen and Corkin (1981) tried to teach patient H.M. to solve the Tower of Hanoi puzzle.5 Patient H.M. worked on the puzzle four times a day for four days, then, after a week’s interval, for another four days. His performance gradually improved until he finally learned the correct sequence. Obviously, he learned and remembered a complex set of responses. However, he was not able to recall that he had worked on the puzzle on previous days, and he expressed surprise at being able to do so well (Carlson, 1986). The conclusion from this case study is clear. Learning is mediated by a number of independent or semi-independent processes, probably residing in different brain locations, and probably interacting differently with other memory and cognitive mechanisms. It is possible that the preexposure phase of LI can, on the one hand, generate an attentional deficit which interferes with developing new associations during the acquisition phase. On the other hand, during the preexposure phase, learning may occur in parallel or alternative channels, but that learning may not be overtly expressed, except under special test conditions (reminder cues), at which spontaneous recovery of learning will occur (LI will be attenuated).
Evidence from the LI literature and from the present study suggest that such recovery is possible, and therefore at least part of the LI phenomenon is derived from a retrieval failure mechanism activated in test. Parallel learning mechanisms can explain contradictions between CAT and retrieval failure theory and can serve as a bridge between the conflicting theories.
Novel pop-out and LI
NPO is defined by Hawley, Johnston and Farnham (1993, 1994) as better/faster detection of a novel target in an array of familiar items, than in an array in which all of the items are novel. In their study, observers viewed arrays composed of strings of letters. Subject were tested on accuracy of string localization after each array presentation. In the present study the NPO effect was measured in a similar way, by RT differences between NPE (novel target among familiar distractors) and NOV (novel array- target and distractors are novel) conditions.
When designing the present experiments, it appeared that NPO might play a significant role in explaining LI. It was suggested that NPO produces an excitatory effect (NPE) while the traditional PE condition produces an inhibitory effect, and that the typical LI effect might be the result of a summation of the two (Lubow & Kaplan, 1996). Results from the present series of studies do not support this hypothesis. The size of the NPO effect was varied across experiments. It was strong in Experiment 1, but was weak or absent in subsequent experiments. The robustness of LI compared to the relative instability of NPO suggests independence of LI and NPO effects. Since LI can be demonstrated independently of NPO effect, future LI research using the within-subject design should omit the NOV condition from test. This will result in a more efficient test, with only three conditions, PE, NPE and reminder. Such a test will be shorter or alternatively will allow for more trials with the critical conditions for LI (PE and NPE).
Since we have reached the conclusion that LI and NPO are independent, the NPO phenomenon becomes irrelevant to the present study of LI. However, we can still ask why NPO was present in Experiment 1 and absent or weak in the other experiments. We can assume that the changes in NPO magnitude were a result of change in the test conditions. In the test phase of Experiment 1, there were seven familiarity conditions. Four of them, PE, NPE, NOV and Reminder, also were present in the other experiments. The additional three test conditions, T->T:N, N:D->D, and N:T->D, appeared only in Experiment 1. As NPO was found in Experiment 1, and as it was absent (or weak) in the other experiments, these three omitted conditions may have been critical for eliciting NPO. In the same way as the Reminder condition was critical for LI, we encounter, again, with a phenomenon (NPO) which we predicted to be effected only from preexposure phase, but which also may be effected by the context of test. In the case of NPO, this is not very surprising, because in contrast to LI procedures, in which the preexposure phase usually precedes test phase with long intervals and trials, NPO procedures are usually “short-term”. For example, Hawley, Johnston and Farnham (1993, 1994), demonstrate NPO in a procedure which consisted of an attended array (preexposure) and probe array (test) which take place within an interval of about one second. Therefore, it is reasonable to assume that inter-test conditions should affect NPO.
What is the critical character of the T->T:N, N:D->D, and N:T->D conditions that were necessary to elicit NPO? (The presence of these conditions in Experiment 1 resulted in a robust NPO effect, and their omission in the other experiments resulted in a weak or absent NPO effect). They all appear to share one attribute. The target-distractor combination in these three conditions consist of one novel component and one familiar component. The NPO effect is based on the relations between a novel component and a familiar component (the novel pops out from the familiar). It is possible that these three conditions served as occasion setters for NPO (i.e., for the relations between NPE and NOV test conditions). If this is true, they would have generated a test environment (context) that caused the subject to respond the NPE and NOV conditions differently than to the same conditions in other test environments. They served as reminders or facilitators for associations from the preexposure phase which were represented by NPE (T->D). This increased the contrast between novel and familiar in test (novel pop-out) and thereby, increased RT differences between NPE and NOV.
However, these are post hoc speculations that should be treated cautiously, since no special manipulation was pre-designed in the present study to explain the variations in NPO magnitude across experiments.
Negative priming and LI
Although negative priming was not identified as one of the phenomena being investigated in the present study, the experimental conditions were such as to allow for the generation of negative priming effects. In a number of ways, negative priming is reminiscent of LI, and the question as to their similarity has been raised in earlier discussions. The most salient component which LI and negative priming share is the performance retardation effect. Trials which are preceded by conflicting information, produce longer RTs than trials which are not preceded in that manner.
However, as will be demonstrated, LI and negative priming, in spite of the similar performance effects, are very different. It will be argued that the LI effect found in the present study is not a result of negative priming. We will also consider a possible interaction between the two phenomena.
Two different sources of negative priming can be identified in the present procedures. The first is derived from preexposure and the second from the test phase. For the first source, one could argue that preexposure is actually a priming procedure that results in probe (test) in a conflicting condition (i.e., PE test condition is actually a probe condition of negative priming). For the second source, the interaction between test trials/conditions could result in a negative priming effect (i.e., a PE test trial/condition preceded by a reminder trials could produce negative priming; also, a reminder test trial preceded by a PE test trial could produce negative priming). The two possible sources of negative priming will be discussed separately.
Is LI “preexposure and test” equivalent to negative priming “prime and probe”?
Traditional negative priming is basically a one-stage procedure. It uses single or few priming trials and the test or probe is usually administered immediately after the prime. As a result, at the time of probe, representation of the prime is still active in short-term memory; encoding of the information into long-term memory in order to obtain the primary effect is not required. LI, on the other hand, is requires a two-stage procedure, with a relatively large number of stimulus preexposure preceding the test phase. Furthermore, the effect is measured in a test that can be administered after considerably longer intervals after preexposure (time and trials) compared to the typical prime-probe negative priming intervals. The effects of such LI preexposures are relatively long lasting, and therefore suggests encoding of at least some preexposure information into long-term memory.
In most reports, negative priming effects are manifested, at the most, for a few seconds and over a few trials (1-6 trials). Priming (equivalent to preexposure) is induced, in most such studies, less than a second before the test (May, Kane and Hasher, 1995; Tipper, 1985; but see DeSchepper and Treisman, 1996). These restrictions are not present in traditional LI studies nor in the present study.
Another critical difference is that of context-specificity. A change of context from preexposure to test severely attenuates LI. Negative priming, on the other hand, endures changes from prime to probe. For example, “negative priming is tied to reusing neither a specific response nor the same response mode across trials… participants responding to target letters either vocally, with a key press, or by alternating between vocal and key press, all show equivalent negative priming… negative priming occurs despite physical changes in stimulus type from prime to test trials, including changes from uppercase to lowercase letters, from pictures to words, across sensory modalities, and from a word or picture to its semantic associates, as from the distractor ‘dog’ to the target ‘cat’…” (May, Kane, and Hasher, 1995). These changes from prime to probe, which have only small, if any, effects on negative priming, would disrupt LI.
Although negative priming is usually a “short-term” phenomenon, a study of DeSchepper and Treisman (1996) has demonstrated a “long-term” effect. The results show that representations of nonsense shapes, formed in a single trial and without attention, can last without decrement up to a month. These unique results suggest that LI and negative priming share similar properties. In spite of the critical differences, mentioned above between the two phenomena, they may share a similar mechanism for negative transfer of information.
Nevertheless, a closer examination of this procedure also can suggest that the negative priming effect of DeSchepper and Treisman (1996) is actually an LI effect. DeSchepper and Treisman’s (1996) primary task (based on Rock & Gutman, 1981 – used two overlapping nonsense shapes. The participant in prime/preexposure is asked to attend to only one of the two), was used also as a preexposure task in the LI study of Lubow and Kaplan (1997, Experiment 1). It is apparent that this task contains a built-in masking task (in LI terms). The attended shape is equivalent to a masking task stimulus (target) and the unattended shape is equivalent to the to-be-tested task stimulus (distractor). A masking task is used in LI studies with adult humans because the preexposed stimulus has to be presented so that the subject does not attend to it. Without the masking task, LI is not obtained. Negative priming has no such requirement. Therefore, in DeSchepper and Treisman (1996), the prime may have functioned as a typical LI preexposure phase presentation and the observed effect in response to the probe after a month, would be actually an LI effect. However, a single trial of preexposure, as used in DeSchepper and Treisman (1996), is not typical to LI, and therefore this suggestion should be treated cautiously. (However, the effect reported by DeSchepper and Treisman (1996) has not been yet replicated, even by studies of Treisman).
To conclude, the similarities between LI and negative priming cannot be ignored and we should consider the possibility that the two phenomena are derived from the same cognitive processes.
Can the LI effect which was found in the present study be
derived from negative priming within the test phase?
One could argue that the LI effect found in the present study might be a negative priming effect resulting from the test phase procedures. For example, if two consecutive trials in test consist of reminder and PE conditions (in any order), they should result in a negative priming effect (target:distractors A:B turn to B:A). Three observations from the present study eliminate such a claim:
An integration between LI and negative priming – suggestion for future research
As already mentioned, in spite of the differences between LI and negative priming, there are still questions about common processes for these two phenomena. One framework for viewing the similarity between LI and negative priming is that of “occasion setting”. In negative priming, the prime creates a setting in which the probe is responded to differently, compared to a setting in which it was not primed. We can treat this setting as a context, which serves as an occasion setter that influences future information processing (negative priming in probe). Similarly, in LI, the preexposure generates a setting/context which affects performance in test. In both, LI and negative priming, there is negative transfer from preexposure/primes to test/probes. The typical negative priming procedure transfers information negatively from one trial to a succeeding trial. The typical LI procedure transfers information negatively from the preexposure phase to the test phase which occurs some time later. Within this framework, LI can be viewed as “long-term” negative priming, and, vice versa, negative priming can be viewed as “short-term” LI.
These two phenomena can be integrated. However, the integration, first requires an examination of a negative priming study by Stadler and Hogan (1996). They generated varieties of positive and negative priming with 360 prime-probe pairs. The probes represented seven conditions of familiarity, similar to those defined in Experiment 1. Unlike in the present study, Stadler and Hogan (1996) did not define the familiarity conditions in terms of the relationship between preexposure and test phases, but rather between trials. Each pair of trials (n-1 and n) consisted of a prime and probe. All conditions were previously primed targets and/or distractors that were presented on the probe trial as either targets or distractors. A novel probe stimulus served as the control. The prediction for D->T:T->D (PE) versus N:T->D (NPE) is identical for LI and negative priming as defined by Stadler and Hogan (1996): slower RTs for PE than for NPE.
It would be of interest to check the interaction between these prime (test trial) and preexposure (phase) effects. This can be done by combining the procedure of Experiments 4, with the procedure of Stadler and Hogan (1996). In the first phase of the combined procedure a standard preexposure phase will be administered (Experiment 4). This preexposure phase will contain homogenous presentations of target and distractors (same target and same distractors will be presented in all preexposure frames). In the second phase, a test procedure will be administered, containing, as in Stadler and Hogan (1996), many prime-probe pairs. Both the effects of preexposure and of prime will be examined only on probe trials. Four test familiarity conditions will be defined for each of these effects. Table 1 defines each of these conditions.
The four within-subject preexposure-effect conditions, and four within-subject prime-effect conditions, generate a 4×4 matrix which defines 16 types of “n trials”. Table 2 and Table 3 illustrate the possible combinations.
The 4×4 matrix sets the stage for generating a variety of predictions. For example, LI magnitude, defined as PE vs. NPE, will be higher if PE is accompanied by negative priming and NPE is accompanied by positive priming, compared when PE is accompanied by positive priming and NPE is accompanied by negative priming.
Preliminary data from pilot studies reveal an interesting pattern of results. PE conditions produced longer RTs than NPE conditions (LI), no matter which primes (n-1 trial) accompanied PE or NPE, i.e., LI was manifested both when PE or NPE were accompanied by negative or positive prime conditions (all combinations). However, a stronger LI effect was found when PE and NPE were accompanied by negative priming and positive priming, respectively, compared to conditions where PE and NPE were accompanied by positive priming and negative priming, respectively.
Such future research on LI and negative priming, which will integrate the two procedures should allow for an independent evaluation of preexposure and priming effects, as well as the interaction between them. More importantly, these data will help to define the similarities and differences in processing mechanisms.
The present dissertation developed a new approach, based on a within-subject visual search procedure, to the measurement of the attentional processes that are believed to underlie LI. Six Experiments provided evidence that stimulus preexposure in a visual search task consisting of a single target amongst many homogenous distractors affects subsequent test phase performance in which targets and distractor vary in their familiarity and status relative to that of the preexposure phase. In addition to LI effect demonstrated in the experiments, it was found that test-task complexity, context constancy from preexposure to test, and reminder cues in test which serve as retrieval cues to the preexposure experience, all affect LI as represented in visual search performance.
The new visual search procedure has recently been used in the study of LI in special populations in which unique LI effects were expected, either attenuated LI or super-LI effects compared to normal controls. Such populations include medicated and non-medicated schizophrenics, Parkinson’s disease patients, anxiety disorder children and adults, and brain-damaged patients (e.g., Lubow, Oren, Laor, and Kaplan, 1998a, 1998b; Lubow, Dressler, and Kaplan, 1998). This new approach allows for a relatively simple, within-subject procedure for collecting LI data, while at the same time addressing a number of theoretical questions and allowing for the possibility of integrating several different empirical phenomena. As such, it is expected to have a considerable impact on the study of normal LI as well as attentional dysfunctions in psychopathological populations.
Tel Aviv, July, 1999.
Ackil, J.E., and Mellgren, R.L. (1968). Stimulus preexposure and instrumental learning. Psychonomic Science, 11, 339.
Arwas, S., Rolnick, A., and Lubow, R.E. (1989). Conditioned taste aversion in humans using motion-induced sickness as the US. Behavior Res. Ther., 27, 295-301.
Bakner, L., Strohen, K., Nordeen, M. and Riccio, D.C. (1991). Postconditioning recovery from the latent inhibition effect in conditioned taste aversion. Physiology and Behavior, 50, 1269-1272.
Baruch, I., Hemsley, D.R., and Gray, J.A. (1988). Differential performance of acute and chronic schizophrenics in a latent inhibition task. Journal of Nervous and Mental Disease, 176, 598-606.
Broadbent, D.E. (1958). Perception and communication. London: Pergamon Press.
Brooks, D.C., and Bouton, M.E. (1993). A retrieval cue from extinction attenuates spontaneous recovery. Journal of Experimental Psychology: Animal, Behavior Processes, 19,77-89.
Bouton, M.E. (1987). Context, encoding, and retrieval in latent inhibition. Unpublished manuscript.
Bouton, M.E. (1993). Context, time, and memory retrieval in the interference paradigms of Pavlovian learning. Psychological Bulletin, 114, 80-99.
Bouton, M.E. (1994). Conditioning, Remembering, and forgetting. Journal of Experimental Psychology: Animal Behavior Processes, 20, 219-231.
Braunstein-Bercovitz, H.and Lubow, R.E. (1998). Attentional involvements in latent inhibition and the implications for attentional dysfunctions in "psychotic – prone" normal subjects. Unpublished Master's Thesis in Psychology, Tel-Aviv University.
Braunstein-Bercovitz, H. and Lubow, R.E. (1997). Latent inhibition as a function of masking task load in low and high schizotypal normals: who is more distractible? Submitted.
Carlson, R.C. (1986). Physiology of Behavior. Boston: Allyn and Bacon, Inc.
Cohen, N.J., and Corkin, S. (1981). The amnestic patient H.M.: learning and retention of a cognitive skill. Society for Neuroscience Abstracts, 7, 235.
Coren, S., and Ward, L.M. (1989). Sensation & Perception. Orlando: Harcourt Brace Jovanovich, Publishers.
Corkin, S., Sullivan, E.V., Twitchell, T.E. and Grove, E. (1981). The amnesic patient H.M.: Clinical observations and test performance 28 years after operation. Society for Neuroscience Abstracts, 7, 235.
Daum, I. (1997). Personal correspondence.
De La Casa, G. and Lubow, R.E. (1995). Latent inhibition in conditioned taste aversion: the role of stimulus frequency and duration and the amount of fluid ingested during preexposure. Neurobiology of Learning and Memory, 64, 125-132.
De La Casa, G. (1998). Personal correspondence.
De La Casa, G. (1999). Personal correspondence.
DeSchepper, B. and treisman, A. (1996). Visual memory for novel shapes: implicit coding without attention. Journal of Experimental Psychology: Learning, Memory, and cognition, 22, 27-47.
Egeth, H.E. (1967). Selective attention. Psychological Bulletin, 67, 41-56.
Epstein, W. (1967). Varieties of perceptual experience. New York: McGraw-Hill.
Garcia, J. & Koeling, R.A. (1967). A comparison of aversions induced by X-rays, toxins and drugs in the rat. Radiation research suppl., 7, 439-450.
Gibbons and Rammsayer (1997). Unpublished manuscript.
Gibson, E.J. (1969). Principles of perceptual learning and development. New York: Appleton-Century-Crofts.
Gibson, E.J., and Walk, R.D. (1956). The effect of prolonged exposure to visually presented patterns on learning to discriminate them. Journal of Comparative and Physiological Psychology, 49, 239-242.
Ginton, A., Urca, G., and Lubow, R.E. (1975). The effects of preexposure to a non attended stimulus on subsequent learning: latent inhibition in adults. Bulletin of the Psychonomic Society, 5, 5-8.
Gray, J.A., Feldon, J., Rawlins, J.N.P., Hemsley, D.R., & Smith, A.D. (1991). The neuropsychology of schizophrenia. Behavioral and Brain Sciences, 14, 1-84.
Gray, N.S., Pickering, A.D., Hemsley, D.R., Dawling, S., and Gray, J.A. (1992). Abolition of latent inhibition by a single 5 mg dose of d-amphetamine in man. Psychopharmacology, 107, 425-430.
Hall, G. (1991). Perceptual and associative learning. Oxford, England: Clarendon Press.
Hawley, K.J., Johnston, W.A., and Farnham, J.M. (1994). Novel popout with nonsense strings: Effects of predictability of string length and spatial location. Perception and Psychophysics, 55, 261-268.
Hochberg, J. (1981). On cognition in perception: perceptual coupling and unconscious inference. Cognition, 10, 127-134.
Hochberg, J. (1982). How big is a stimulus? In Beck, J. (Ed.), Organization and representation in perception. Hillside N.J: Lawrence Erlbaum.
Holland, P.C. (1992). Occasion setting in Pavlovian conditioning. In: Median, D.L. (Ed.). The Psychology of Learning and Motivation (Vol. 28 p.69-125). San Diego, CA: Academic Press.
Hull, C.L. (1943). Principles of behavior. New york: Appleton-Century-Crofts.
Johnston, W.A., Hawley, K.J., Plewe, S.H., Elliott, J.M.G. and DeWitt, M.J. (1990). Attention capture by novel stimuli. Journal of Experimental Psychology: General, 119, 397-411.
Kahaneman, D. (1973). Attention and Effort. New Jersy: Prentice-Hall.
Kaplan, O. (1992). The context effect: application of latent inhibition and perceptual learning paradigms to consumer attitudes to products. Unpublished Master's Thesis in M.B.A., Tel-Aviv University.
Kasprow, W.J., Catterson, D., Schachtman, T.R., and Miller, R.R. (1984). Attenuation of latent inhibition by post-acquisition reminder. Quarterly Journal of Experimental Psychology, 36, 53-64.
Kraemer, P.J., and Ossenkopp, K.P. (1986). The effects of flavor preexposure and test interval on conditioned taste aversions in rats. Bulletin of the Psychonomic Society, 24, 219-221.
Kraemer, P.J, Hoffman, H., and Spear, N.E. (1988). Attenuation of the CS-preexposure effect after a retention in preweanling rats. Animal Learning and Behavior, 16, 185-190.
Kraemer, P.J., Randall, C.K., and Carbary, T.J. (1991). Release from latent inhibition with delayed testing. Animal Learning and Behavior, 19, 139-145.
Kraremer, P.J., and Roberts, W.A. (1984). The influence of flavor preexposure and test interval on conditioned taste aversion in the rat. Learning and Motivation, 15, 259-278.
Kraemer, P.J., and Spear, N.E. (1992). The effects of nonreinforced stimulus exposure on the strength of a conditioned taste aversion as a function of retention interval: do latent inhibition and extinction involve a shared process. Animal Learning and Behavior, 20, 1-7.
Krose, B.J.A., and Julesz, B. (1990). Automatic or voluntary allocation of attention in visual search task. In: Brogan, D. (Ed.), Visual search. London: Taylor & Francis Ltd.
Lovibond, P.F., Preston, G.C., and Mackintosh, N.J. (1984). Context specificity of conditioning, extinction, and latent inhibition. Journal of Experimental Psychology: Animal Behavior Processes, 10, 360-375.
Lubow, R.E. (1989). Latent inhibition and conditioned attention theory. Cambridge: Cambridge University Press.
Lubow, R.E. (1992). Personal correspondence.
Lubow, R.E. and Josman, Z. E. (1993). Latent inhibition deficits in hyperactive children. Journal of Child Psychology and Psychiatry and Allied Disciplines, 34, 959-973.
Lubow, R.E., and Kaplan, O. (1997). Visual search as a function of type of prior experience with target and distractor. Journal of Experimental Psychology: Human perception and performance, 23, 14-24.
Lubow, R.E., Dressler, R. & Kaplan, O. (1998). Visual search in de-novo Parkinson's Disease patients and normal controls. Unpublished manuscript to be submitted.
Lubow, R.E., Oren, H, Laor, N, & Kaplan, O. (1998a). Visual search in schizophrenic: latent inhibition and novel pop out effects. Unpublished manuscript to be submitted.
Lubow, R.E., Oren, H, Laor, N, & Kaplan, O. (1998b). Visual search in normal and anxiety disorder children and adolescents. Unpublished manuscript to be submitted.
Lubow, R.E., and Gewirtz, J.C. (1995). Latent inhibition in humans: Data, theory, and implications for Schizophrenia. Psychological-Bulletin, 117, 87-103.
Lubow, R.E., and Moore, A.U. (1959). Latent inhibition: the effect of non reinforced preexposure to the conditioning stimulus. Journal of Comparative and Physiological Psychology, 52, 415-419.
Lubow, R.E., Rifkin, B., and Alek, M. (1976). The effects of context: the relationship between stimulus preexposure and environmental preexposure determines subsequent learning. Journal of Experimental Psychology: Animal Behavior Processes, 2, 38-47.
Lubow, R.E., Schnur, P. and Rifkin, B. (1976). Latent inhibition and conditioned attention theory. Journal of Experimental Psychology: Animal Behavior Processes, 2, 163-174.
Lubow, R.E., Weiner, I., and Schnur, P. (1981). Conditioned attention theory. In: G.H. Bowere (Ed.), The Psychology of Learning and Motivation (Vol. 15 p.1-49), New-York: Academic Press.
Machworth, J.A. (1970). Vigilance and habituation. London: Penguin Press.
Mackintosh, N.J. (1975). A theory of atte: Variations in the associability of stimuli with reinforcement. Psychological Review, 82, 276-298.
May, C.P., Kane, M.J., and Hasher, L. (1995). Determinants of Negative Priming. Psychological Bulletin, 118, 35-54.
McGeoch, J.A. (1932). Forgetting and the law of disuse. Psychological review, 39, 352-370.
Miller R.R., Kasprow, W.J. and Schachtman, T.R. (1986). Retrieval variability: Sources and consequences. American Journal of psychology, 99, 145-218
Milner, B., Corkin, S., and Teuber, H.L. (1968). Further analysis of the hippocampal amnesic syndrome: 14 year follow-up study of H.M. Neuropsychology, 6, 317-338.
Milner, B. (1970). Memory and the temporal regions of the brain. In Pribram, K.H. and Broadbent, D.E. (Ed.). Biology of memory. New York: Academic Press.
Muller, H.J., Humphreys, G.W., Quinlan, P.T., and Riddoch, M.J. (1990). Combine-feature coding in the form domain. In: David Brogan (1990), Visual Search, chp. 5. London: Taylor & Francis.
Musem, G. and Triesman, A. (1990). Implicit and explicit memory for visual patterns. Journal of Experimental psychology: Learning, Memory and Cognition, 16, 127-137.
Nachman, M. (1970). Learned taste and temperature aversions due to lithium chloride sickness after temporal delays. Journal of comparative and physiological psychology, 73, 22-30.
Oswalt, R.M. (1972). The relationship between level of visual pattern difficulty during rearing and subsequent discrimination in rats. Journal of comparative and Physiological Psychology, 81, 122-125.
Pavlov, I.P. (1927). Conditioned reflexes. London: Oxford University Press.
Pearce, J.M., and Hall, G. (1980). A model of Pavlovian learning: Variations in the effectiveness of conditioned but not unconditioned stimuli. Psychological Review, 87, 532-552.
Posner, M.I., and Stephen, J.B. (1971). Components of attention. Psychological Review, 78, 391-408.
Rescorla, R.A. (1971). Summation and retardation tests of latent inhibition. Journal of Comparative and Physiological Psychology, 75, 77-81.
Rescorla, R.A., and Cunningham, C.L. (1978). Contextual learning in Pavlovian conditioning. In: Balsam, P.D., and Tomie, A. (Eds.). Context and Learning. 23-56. Hillsdale, N.J.: Erlbaum.
Rescorla, R.A. (1986). Extinction of facilitation. Journal of Experimental Psychology: Animal Behavior Processes, 12, 16-24
Revusky, S.H.& Bedarf, E.W. (1967). Association of illness with prior ingestion of novel foods. Science, 155, 219-220.
Robins, S.J. (1990). Mechanisms underlying spontaneous recovery in autoshaping. Journal of Experimental Psychology: Animal Behavior Processes, 16, 235-249.
Rock, I. and Gutman, D. (1981). The effect of inattention on form perception. Journal of Experimental Psychology: Human Perception and Performance, 7, 275-285.
Roter and Lubow (1998). Personal correspondence.
Salzman, A.,Hadar, U., Korczyn, A., and Lubow, R.E. (1994). Latent inhibition in Parkinson patients and elderly subjects. Unpublished manuscript, Department of Psychology, Tel-Aviv University.
Shiffrin, R.M. and Schneider, W. (1977). Controlled and automatic information processing: II. Perceptual leaning, automatic attending, and a general theory. Psychological Review, 84, 127-190.
Stadler, M.A., and Hogan, M.E. (1996). Varieties of positive and negative priming. Psychonomic Bulletin and Review, 3, 87-90.
Tipper, S.P. (1985). The negative priming effect: Inhibitory effects of ignored primes. Quarterly Journal of Experimental Psychology, 37, 571-590.
Tolman, E.C. (1932). Purposive behavior in animals and men. New York: Appleton-Century.
Thomas, D.R., and Sherman, L (1986). An assessment of the role of handling cues in spontaneous recovery after extinction. Journal of the Experimental Analysis of Behavior, 46, 305-314.
Tulving, E. (1974). Cue-dependent forgetting. American Scientist, 62, 74-82.
Triesman, A. (1969). Strategies and models of selective attention. Psychological Review, 76, 282-299.
Wagner, A.R. (1976). Priming in STM: An information processing mechanism for self generated or retrieval-generated depression in performance. In: T.J. Tighe and R.N. Leaton (Eds.), Habituation: Perspectives from Child Development, Animal Behavior, and Neurophysiology, 95-128. Hillsdale, N.J.: Lawrence Erlbaum.
Wangner, A.R. (1978). Expectancies and the priming of STM. In S.H. Hulse, H. Fowler, and Honig, W.K. (Eds.), Cognitive processes in animal behavior, 177-209. Hillsdale, N.J.: Lawrence Erlbaum.
Wagner, A.R. (1981). SOP: A model of automatic memory processing in animal behavior. In: Spear, N.E. and Miller R.R. (Eds.), Information Processing in Animals: Memory Mechanisms, 5-47. Hillsdale, N.J.: Lawrence Erlbaum.
Zajonc, R.B. (1968). Attitudinal effects of mere exposure. Journal of Personality and social Psychology Monographs, 9, 1-27.
Zalstein-Orda N. and Lubow R.E. (1995). Context control of negative transfer induced by preexposure to irrelevant stimuli: Latent inhibition in humans. Journal of Learning and Motivation, 26, 11-28.
* The present dissertation is a result of a long research process which was encountered with many theoretical and technical issues. These issues will be presented and discussed in each of the following experiments. Of course, it is not possible to discuss these issues before presenting the data. To make up with the absence of these issues in the introduction section, each experiment in the results section has an introduction and discussion section. At the end of the results section there is a general discussion section which summarize the whole research process.
5 This puzzle consists of three wooden spindles, one of which contains five disks arranged according to size, the smallest on top. The task is to place the disks, still arranged according to size, on another spindle, moving one disk at a time. There is one restriction: a disk can be placed on an empty position or on top of a larger disk, but never on top of a smaller one. The strategy for solving this task is not immediately obvious but must be learned through trial and error (such tasks are very common in LI human studies).