Semantic Memory and Priming Research Paper

The plan of this research paper is as follows: In the first section, we review models of semantic memory proposed in the 1960s and 1970s and the major empirical findings that were used to test these models. We also summarize two contemporary models of semantic memory, distributed network models and high-dimensional spatial models. In the second section of the paper, we examine semantic priming. We review the most influential models of semantic priming and then summarize empirical developments, focusing in particular on issues that have turned out to be important for testing models of semantic priming. In the final section of the paper, we look at repetition priming, reviewing both models and major issues and findings. We close with a brief summary of our major conclusions.

Semantic Memory

Semantic memory refers to our knowledge about language and facts about the world; it can be thought of as a mental dictionary, encyclopedia, and thesaurus all rolled into one (e.g., E. E. Smith, 1978; Tulving, 1972). A defining characteristic of semantic memories is that we, as introspective observers, do not know where they came from; they are not represented in terms of specific times and places. Semantic memory has traditionally been contrasted with episodic memory (e.g., Tulving, 1983). Episodic memory refers to our knowledge that is tagged temporally or spatially, or identified in some way in terms of personal experiences. Although there are reasons to believe that semantic memory and episodic memory are not independent systems (e.g., McKoon, Ratcliff, & Dell, 1986), the distinction has been extremely influential in the field of memory and is useful for organizing memory phenomena, tasks, and models.

A complete theory of semantic memory should be able to explain the following phenomena (e.g., E. E. Smith, 1978): First, a theory of semantic memory should explain how the meanings of words are mentally represented. It might specify, for example, that meaning is represented as a collection of features, some of which are essential and others of which are just typical (e.g., for bird, animate, and can fly, respectively). Second, it should be able to explain how the meanings of individual words can be combined to form more complex units. How, for example, is the meaning of a simple noun combination, such as pet bird, constructed from the meanings of its constituents, pet and bird? Third, the theory should specify the permissible inferences that can be made from word and sentence meanings. What can you infer about a grampus if you know that it is a mammal? This goal is, of course, closely tied to the first. Fourth and finally, a theory of semantic memory should explain the connection between word meaning and the world, between semantic representations and perceptual systems. For example, it should explain how we recognize an object from a description, or describe an object based on perceptual input (e.g., vision, taction, etc.).

As a matter of history, theories of semantic memory have dealt primarily with the first goal, specifying how word meanings are mentally represented. There has been a fair amount of research on how word meanings are combined, but it has been carried out under the guise of investigations of concepts and categorization. Very little attention has been given to the third and the fourth goals by cognitive psychologists. Our goal in this section of the research paper is to review some of the major theoretical and empirical developments in the field of semantic memory. We begin by summarizing the models developed during the late 1960s and the 1970s, the golden age of semantic memory research. We then turn to a brief review of some of the major empirical challenges posed during that time. We close with a brief review of recent models of semantic memory.

Early Models of Semantic Memory

The early models of semantic memory were of three basic types:networkmodels(e.g.,Collins&Loftus,1975;Collins& Quillian, 1969; Glass & Holyoak, 1974; Quillian, 1967); settheoretic models (D. E. Meyer, 1970); and feature models (e.g.,McCloskey&Glucksberg,1979;E.E.Smith,Shoben,& Rips, 1974). Two of these models turned out be extremely influential: the spreading-activation theory of Collins and Loftus (1975) and the feature-comparison theory of E. E. Smith et al. (1974). We focus our attention on these two models. For a comprehensive review of the other models, consult E. E. Smith (1978).

Spreading-Activation Theory of Semantic Processing

The spreading-activation theory of semantic processing proposed by Collins and Loftus (1975) is an elaboration of the hierarchical network model proposed by Quillian and Collins (e.g., Collins & Quillian, 1969; Quillian, 1967).Aunique feature of the model, at least in the context of psychological models of semantic memory, is that it distinguishes knowledge of the meanings of concepts from knowledge of their names.

The conceptual network is organized according to semantic similarity. Concepts are assumed to be represented as nodes in a network. The more properties two concepts have in common, the more links that exist between the two nodes. For example, car and truck would have many links between them, whereas car and apple would have few links. In the original hierarchical network model, several types of links were distinguished (e.g., superordinate and subordinate, modifiers, disjunctive sets, etc.).This rich array of link types allowed the model to account for a wide variety of semantic decisions (e.g., Quillian, 1969). However, the different link types did not play an important role in the elaborated theory.

The names of concepts are stored in a lexical network organized according to phonemic similarity. Thus, for example, several links would exist between the nodes for car and bar, but no links would exist between the nodes for car and bus. Each node in the lexical network is connected to at least one node in the conceptual network.

The fundamental retrieval mechanism is spreading activation. Concepts are activated by being mentally processed in some manner; for example, thinking about or seeing apples would activate the corresponding concept in semantic memory. Activation spreads from a concept along links throughout the network and decays with distance in the network; that is, the farther the activation spreads, in terms of number of links traversed, the less arrives at the destination. Activation also requires more time to spread greater distances. Activation is released from a concept as long as it is processed, but only one concept can be actively processed at any one time, and therefore only one concept can be a source of activation. Activation gradually decays with time if no concepts in the network are being processed.

Several ancillary processing assumptions are made to handle particular semantic judgments. One such assumption is that people can control whether to activate the conceptual network or the lexical network. For instance, an individual could try to think of exemplars of bird, which involves activating the conceptual network, or think of words that sound like bird, which involves activating the lexical network. Another assumption is that semantic decisions, such as verifications of member-category and property statements (e.g., A robin is an animal and A robin has feathers, respectively), are made by accumulating positive and negative evidence until a positive or a negative criterion is reached. The evidence consists of various kinds of connections that are found during the memory search. For example, for a member-category statement, such as A robin is an animal, the superordinate connections from robin to bird and from bird to animal would count as positive evidence. These evidence accumulation processes are very similar to the processing assumptions of the feature comparison theory later described.

One of the longest lasting impacts of Collins and Loftus’s (1975) model came from its ability to provide an elegant explanation of semantic priming; indeed, this model became the canonical model of semantic priming. According to this model, processing of a prime word causes activation to spread from the prime throughout the conceptual network. More activation will accumulate at concepts close to the prime than at concepts far from the prime. This residual activation then facilitates the semantic decision on the target word. For example, because bird and robin are closer in memory than are dog and robin, more activation accumulates at robin when bird is the prime than when dog is the prime, and decision times are correspondingly faster.

Feature-Comparison Theory

The feature-comparison theory (e.g., E. E. Smith et al., 1974) has two major sets of assumptions, those concerning the representation of word meaning and those concerning the processing of word meaning.

The meaning of a word is represented by a set of semantic attributes or features. The features vary continuously on a scale of “definingness”: At one end of the scale are features that are essential to the word’s meaning; at the other end of the scale are features that are only characteristic of the concept. For example, the concept mammal might include as defining features the facts that mammals are animate, have mammary glands, and nurse their young, and as characteristic features the facts that mammals give birth to live young, have four limbs, and live on land.

It is assumed in the model that verification of a statement, such as A dog is an animal, involves a two-stage process. In the first stage, a global index of meaning similarity is computed by matching all of the features in the subject and the predicate. If this index of similarity exceeds an upper criterion (e.g., A dog is an animal), a rapid true decision is made, and if it falls below a lower criterion (e.g., A dog is furniture), a rapid false decision is made. However, if the similarity index is intermediate in value (e.g., A dog is a quadruped), the defining features of the predicate are compared to those of the subject. If all match, the statement is true, whereas if any mismatch, the statement is false.

The basic predictions of the model rely on the assumption that response latencies are faster for statements that can be verified by the first stage than for statements that require both stages. For true statements, the model predicts that statements will be verified faster, on the average, if the subject and the predicate are highly semantically related than if they are not highly related. The reason is that the global index of meaning similarity is more likely to exceed the upper criterion for semantically related subjects and predicates, and therefore processing of the statement is more likely to engage only the first stage. E. E. Smith et al. (1974) assumed that typicality ratings and association norms were reflections of featural similarity between concepts. Hence, the model predicts, in particular, that true statements will be verified faster if the subject is a typical exemplar than if it is an atypical exemplar of the predicate category (e.g., A robin is a bird vs. A penguin is a bird).

For false statements, the more similar the subject and the predicate, the less likely the statement is to fall below the lower criterion. Therefore, similar false statements (e.g., A bat is a bird) should be more likely to engage the second stage of processing, and so take longer to reject, than dissimilar false statements (e.g., A robin is furniture). Although this prediction has been confirmed (e.g., E. E. Smith et al., 1974), it has also been disconfirmed for certain types of false statements (Holyoak & Glass, 1975), as discussed below.

Major Issues and Findings

Collins and Loftus’s (1975) spreading activation theory is sufficiently complex that it is probably unfalsifiable (but see the section below on semantic priming). In contrast, both the hierarchical network model and the feature comparison model made strong assumptions about how meanings were represented and processed and, therefore, made testable predictions about performance in semantic decision tasks. In the following paragraphs, we summarize two lines of research that were influential in testing these models.

Associative Strength and Typicality

Collins and Quillian’s (1969) hierarchical network model made two crucial assumptions: First, noun concepts were assumed to be stored in a hierarchy determined by the logic of class relations. A concept was stored closer to its immediate superordinates than to its more distant ones. For example, robin was represented as a bird, and bird was represented as an animal, but robin was not directly represented as an animal. The second assumption, which was referred to as “cognitive economy” (Conrad, 1972), held that properties were stored at the highest possible semantic level to which they applied. Continuing the example, feathered would be stored with bird but not with robin, because all birds are feathered, whereas can fly would be stored with robin but not with bird, because robins can fly but not all birds can fly.

These assumptions generate two testable predictions: First, member-category statements should be verified faster if the subject is paired with an immediate superordinate, as in A robin is a bird, than if the subject is paired with a more distant superordinate, as in A robin is an animal. Second, property statements should be verified faster if the subject is paired with a property stored with it, as in A bird has feathers, than if the subject is paired with a property stored at a higher semantic level, as in A bird eats. Both of these predictions were confirmed (e.g., Collins & Quillian, 1969, 1972).

However, the hierarchical network model soon ran into trouble. Conrad (1972) observed that Collins and Quillian (1969) might have confounded hierarchical distance and associative strength. She argued, for example, that A bird has feathers might have been verified faster than A bird eats because bird and feathered are more highly associated than are bird and eats, not because of a difference in network distance. Conrad independently manipulated (a) the hierarchical distance between concepts and their properties, as determined by the assumptions of hierarchical storage and cognitive economy, and (b) the associative strength between concepts and their properties, as measured by association norms. She found that verification time decreased as associative strength increased, but it was insensitive to hierarchical distance. Rips, Shoben, and Smith (1973) also found that some member-category statements involving immediate superordinates took longer to verify than those involving distant superordinates. For instance, A dog is a mammal took longer to verify than A dog is an animal. This result conflicts directly with the hierarchical storage assumption.

Subsequent studies (e.g., E. E. Smith et al., 1974) showed that the critical determinant of decision times was the strength of semantic or associative relation between the subject and the predicate. These studies also demonstrated that typical exemplars of a category (e.g., robin of bird) were verified faster than were atypical exemplars (e.g., chicken). The hierarchical network model did not have mechanisms to explain such findings.

False Statements and Similarity

As described earlier, one of the predictions of the feature comparison theory is that false statements containing similar concepts should be more difficult to reject than false statements containing dissimilar concepts. Holyoak and Glass (1975) showed that this prediction was violated for two kinds of statements. In one kind, similar statements expressed contradictions that were assumed to be directly represented in memory (e.g., All fruits are vegetables, Some chairs are tables), whereas less similar statements did not (e.g., All fruits are flowers, Some chairs are beds). In the other kind of statement, similar statements, but not dissimilar ones, could be disconfirmed by the retrieval of a salient counterexample (e.g., canary for All birds are robins). The importance of these findings is that they indicate that different kinds of evidence can be used to make semantic decisions. This conclusion does not bode well for models, such as the feature comparison theory, in which a single source of information is the basis of all semantic judgments.

In a comprehensive investigation of the processing of false statements, Ratcliff and McKoon (1982) used a responsesignal procedure to trace the time course of processing. An important finding was that performance on category-member statements (e.g., Abirdisarobin) was nonmonotonic: Earlyin processing, there was an increasing tendency to respond true to these false statements, but, later in processing, there was an increasing tendency to respond correctly. This result indicates that, later in processing, new information became available or a second stage of processing was invoked. The nonmonotonicity is problematic for the network models of semantic memory, but it seems to offer support for feature comparison theory. However, at all points in processing, including the very earliest stages, subjects were more likely to respond true to member-category statements (e.g., A robin is a bird) than to category-member statements, even though these statements have equal amounts of overall feature overlap. This finding is not consistent with feature-comparison theory.

Contemporary Approaches to Semantic Memory

Research on semantic memory flourished in the late 1960s and 1970s but was already languishing in the early 1980s. Cognitive psychologists did not lose interest in semantic memory phenomena but, rather, migrated to more specialized programs of research, such as word recognition, language comprehension and production, and concepts and categories. The models developed to account for these phenomena were necessarily more focused than were the original models of semantic memory. In this section of the research paper, we take a quick look at two more recent approaches to understanding knowledge representations.

Distributed Network Models

Distributed network models have a long history (e.g., Hebb, 1949; Rosenblatt, 1962), but they did not become influential in cognitive psychology until the mid-1980s (e.g., McClelland & Rumelhart, 1986; Rumelhart & McClelland, 1986). The development and investigation of distributed network models has become a gigantic enterprise. Our goal will be to summarize the most important characteristics of these models, especially as they apply to semantic memory. According to distributed network models, concepts are represented as patterns of activation across a network of densely interconnected units. Similar concepts are represented by similar patterns of activation. The units can be thought of as representing aspects of the object or event being represented. These aspects, however, need not be nameable or correspond in any obvious way to the features people might list in a description of the entity. Indeed, a traditional feature, such as has wings, might itself be a pattern of activation over a collection of units.

Units are typically organized into modules, which correspond to sets of units designed to represent a particular kind of information (e.g., verbal vs. visual) or to accomplish a particular information processing goal (e.g., input vs. output). For example, Farah and McClelland’s (1991) model of semantic memory impairment has three modules corresponding to verbal inputs, to visual inputs, and to semantic representations (which are further subdivided into visual units and functional units). Units within a module are richly interconnected with each other, and units in different modules may or may not be connected depending on the architecture of the model. For example, in Farah and McClelland’s model, visual input units and verbal input units are connected to semantic representation units but not to each other.

Presenting a stimulus to the network causes an initial pattern of activation across the units, with some units more active than others. This pattern changes as each unit receives activation from the other units to which it is connected.Astable pattern of activation eventually appears across the units. The particular pattern instantiated across a set of units in response to an input, such as seeing an object or hearing a word, is determined by the weights on the connections between the units. Knowledge is therefore encoded in the weights, which constitute the long-term memory of the network.

The feature of distributed network models that may explain more than any other their continuing influence is that they learn. A network can be trained to produce a particular output, such as the meaning of a word, in response to a particular input, such as the orthographic pattern of the word. Training involves incrementally adjusting the weights between units so as to improve the ability of the network to produce the appropriate output in response to an input.

Another important characteristic of distributed network models is that their performance can decay gracefully with damage to the network. This characteristic is a result of having knowledge distributed across many connection weights in the network. For example, even with up to 40% of its visual semantic memory units destroyed, Farah and McClelland’s (1991) model was able to correctly associate names and pictures more than 85% of the time.

Distributed network models have been applied to many human behaviors that depend on information traditionally represented in semantic memory, including acquisition of generic knowledge from specific experiences (e.g., McClelland & Rumelhart, 1985), word naming and lexical decision (e.g., Kawamoto, Farrar, & Kello, 1994; Seidenberg & McClelland, 1989), impairments in reading and the use of meaning after brain damage (e.g., Farah & McClelland, 1991; Hinton & Shallice, 1991; Plaut, McClelland, Seidenberg, & Patterson, 1996), and (as discussed later) semantic priming. Although these models have had their critics (e.g., Besner, Twilley, McCann, & Seergobin, 1990; Fodor & Pylyshyn, 1988), their influence on the science of memory has been, and promises to remain, enormous.

High-Dimensional Spatial Models

The idea that concepts can be represented as points in space, such that the dimensions of the space correspond to important dimensions of meaning, has a long history (e.g., Osgood, Suci, & Tannenbaum, 1957). This idea has recently been resurrected in two models of the acquisition and representation of word meaning.

Hyperspace Analog to Language (HAL). HAL (e.g., Burgess & Lund, 2000) is a spatial model of meaning representation in which concepts are represented as points in a very high dimensional space. The semantic similarity between concepts is represented by the distance between corresponding points in the space. As a result of the methodology used, meanings of concepts are represented in terms of their relations to other concepts.

The methodology involves tracking lexical co-occurrences within a 10-word moving window that slides across a corpus of text. The corpus includes approximately 300 million words taken from Usenet newsgroups containing English text. HAL’s vocabulary consists of the 70,000 most frequently used symbols in the corpus.About half of these symbols have entries in the standard Unix dictionary; the remainder includes nonwords, misspellings, proper names, and slang. For ease of exposition, we refer to the 70,000 symbols as words. The methodology therefore produces a 70,000 × 70,000 matrix of co-occurrence values.

The co-occurrence matrix is constructed so that entries in each row specify the weighted frequency of co-occurrence of the row word and the words that preceded it in the window; entries in each column specify the weighted frequency of cooccurrence of the column word and the words that followed it in the window. Words that are closer together in the moving window get larger weights. Contiguous words receive a weight of 10; words separated by one intervening word receive a weight of 9; and so forth.

The meaning of a word is captured in the 140,000element vector obtained by concatenating the row and the column vector for that word. Each vector can be thought of as a point in a 140,000-dimensional space. The similarity in meaning between two words is defined as the Euclidean distance between their corresponding points in the space. An important property of HAL is that two words (e.g., street and road) can have very similar meanings because they occur in similar contexts and, hence, have similar meaning vectors, not because they appear frequently in the same sentence (cf. McKoon & Ratcliff, 1992).

HAL is a structural model of meaning and has no processing architecture. Hence, most of the evidence on the model consists of qualitative demonstrations or correlations between indices generated by the model and human behavior. For example, when distances between word vectors are computed and submitted to multidimensional scaling, the resulting scaling solutions indicate that words are grouped into sensible categories (e.g., Burgess & Lund, 2000). Other experiments have shown that interword distances computed in HAL predict priming in lexical decision, to a reasonable approximation (e.g., Lund, Burgess, & Audet, 1996).

Latent Semantic Analysis (LSA). The overarching goal of the LSA model (e.g., Landauer, 1998; Landauer & Dumais, 1997) is to explain Plato’s paradox: Why do people appear to know so much more than they could have learned from the experiences they have had? Like HAL, LSA is a high-dimensional spatial model of meaning representation. Concepts in LSA are represented by vectors in a space of approximately 300 dimensions. Similarities between meanings of concepts are represented by cosines of angles between vectors.

The input to LSA is a matrix in which rows represent types of events and columns represent contexts in which instances of the events occur. In many applications, for example, the rows correspond to word types and the columns correspond to samples of text (e.g., paragraphs) in which instances of the words appear. Each cell in the matrix contains the number of times that a particular word type appears in a particular context. This matrix is analyzed using singular value decomposition (SVD), which is similar to factor analysis. This analysis allows event types and contexts to be represented as points or vectors in a high-dimensional space. In this new representation, the similarities between any pairs of items can be computed.

In one specific implementation, samples of text were taken from an electronic version of an encyclopedia containing 30,473 articles. From each article, a sample was taken consisting of the first whole text or 2,000 characters, whichever was less. The text data were placed in a matrix of 30,473 columns, each representing a text sample, and 60,768 rows, each representing a word that had appeared in at least two samples. The cells in the matrix contained the frequency with which a word appeared in a particular sample. After transforming the raw cell frequencies, the matrix was submitted to SVD and the 300 most important dimensions were retained. Thus, each word and each context could be represented as a vector in a 300-dimensional space.

LSA has been applied to a varied set of problems. In one application, the model’s word knowledge after training was tested using items from the synonym portion of the Test of English as a Foreign Language (TOEFL). Each problem consisted of a target word and four answer options from which the test taker is supposed to choose the one with the most similar meaning to the target. The model’s choices were determined by computing cosines between vector representations of the target words in each item and vector representations of the answer options, and choosing the option with the largest cosine. The model performed as well as applicants to U.S. colleges from non-English speaking countries, getting 64.4% correct.

Another application of the model simulated the acquisition of vocabulary by school-aged children. The model gained vocabulary at about the same rate as do seventh-grade students, approximately 10 words per day. This rate greatly exceeds learning rates that have been obtained in experimental attempts to teach children word meanings from context. An important finding in this analysis was that LSA’s learning of vocabulary relies heavily on indirect learning: The estimated direct effect of reading a sample of text (e.g., a paragraph) on knowledge of words in the sample was an increase of approximately 0.05 words of total vocabulary, whereas the indirect effect of reading a sample of text on words not contained in the sample was an increase of approximately 0.15 words of total vocabulary. Put another way, approximately three fourths of LSA’s vocabulary gain from reading a passage of text was in words not even present in the paragraph. This finding helps to explain, according to Landauer and Dumais (1997), why people can have more knowledge than appears to be present in the information to which they have been exposed.

Summary

The first models of semantic memory appeared in the late 1960s, and by the mid-1970s at least half a dozen comprehensive models had been proposed. The two most influential models were the network model proposed by Quillian, Collins, and Loftus (e.g., Collins & Loftus, 1975; Collins & Quillian, 1969; Quillian, 1967) and the feature-comparison model proposed by E. E. Smith et al. (1974). These models became, and largely remain, the canonical models of semantic memory. Although these early models are no longer considered to be viable accounts of semantic memory, they remain influential because they provide useful ways of conceptualizing and categorizing memory phenomena.

Distributed network models offered an entirely different way of thinking about knowledge representations. In traditional models of semantic memory, concepts were represented by localized nodes or features, and the relations between concepts were either stored in the links (network models) or computed on the fly (feature models). In distributed network models, however, concepts are represented by patterns of activation across many units, which participate in representing other concepts, and knowledge about the relations between concepts is represented across many connection weights, which participate in representing other relations. There is no indication that the influence of these models is flagging.

High-dimensional spatial models also use distributed representations. In these models, however, the meaning of a concept is given by the company it keeps, in written and (presumably) spoken language. Concepts are similar to the extent that they are used in similar contexts. A virtue of these models is that they demonstrate how knowledge can be acquired from specific experiences. A significant challenge for the developers of these models will be to incorporate processing architectures that will allow the models to be subjected to rigorous testing. It remains to be seen how influential these high-dimensional spatial models will turn out to be.

Semantic Priming

Priming is an improvement in performance in a cognitive task, relative to an appropriate baseline, as a function of context or prior experience. Semantic priming refers to the improvement in speed or accuracy to respond to a stimulus when it is preceded by a semantically related or associated stimulus relative to when it is preceded by a semantically unrelated or unassociated stimulus (e.g., cat–dog vs. table–dog; D. E. Meyer & Schvaneveldt, 1971). The stimulus to which responses are made is referred to as thetarget, and the preceding stimulus is referred to as the prime. The other kind of priming examined in this research paper is repetition priming, which refers to an improvement in speed or accuracy to respond to the second (or subsequent) occurrence of a stimulus relative to the first occurrence of the stimulus. Semantic and repetition priming are probably caused by different mechanisms or by different processing stages (e.g., Durgunoglu, 1988), but because they have been so influential in the study of human memory, we review both areas of research in this research paper.

The semantic in semantic priming implies that priming is caused by relations of meaning, as exist, for instance, between the concepts dog and goat (mammals, domesticated, have fur, etc.). In fact, the term has also been used to refer to priming caused by a mixture of semantic and associative relations, as exist between the concepts dog and cat. These concepts are semantically related, but in addition, if people generate associates to dog, they list cat with high frequency (and vice versa). In contrast, goat almost never comes up as an associate of dog. Consistent with usage in the field, we shall use semantic priming to refer to both kinds of priming, unless we need to distinguish the two (as in the section “Associative Versus Pure Semantic Priming”).

Models of Semantic Priming

Spreading Activation Models

Spreading activation was first incorporated into a model of memory by Quillian (1967); this model was elaborated and extended by Collins and Loftus (1975), as described previously. Spreading activation models were also proposed by Anderson (1976, 1983). Although these models differ in several important ways, they share three fundamental assumptions: (a) Retrieving an item from memory amounts to activating its internal representation; (b) activation spreads from a concept to associated concepts; and (c) residual activation accumulating at a concept facilitates its subsequent retrieval. For example, the visual presentation of a word, such as lion, activates its internal representation. This activation spreads to associated concepts, such as tiger. If the word tiger appears soon after the word lion, it can be identified more quickly than normally because it is already partially activated.

Although Collins and Loftus’s (1975) model and Anderson’s (1983) ACT* model are similar, they differ in important ways. The Collins and Loftus model (as well as Anderson’s, 1976, model) assumes that activation takes time to spread from one concept to another. This mechanism is used to explain the effects of hierarchical network distance on verification time. ACT*, in contrast, assumes that activation spreads extremely quickly, reaching asymptote in as little as 50 ms. Effects of network distance are attributed to differences in asymptotic activation levels. Another difference is that Collins and Loftus’s model assumes that activation continues to spread (for a while) even when a concept is no longer being processed. In ACT*, however, activation decays very rapidly, within 500 ms,whena concept ceases to be a source of activation. Finally, the Collins and Loftus model assumes that only one concept can be a source of activation at a time, whereas ACT* assumes that the number of possible sources is limited only by the capacity of attention.

The accounts of semantic priming in the two models are really quite different. In the Collins and Loftus model, the prime sends activation to the target, and the target can be in a preactivated state even though the prime is no longer being processed. In ACT*, however, both the prime and the target must be sources of activation—both must be objects of attention—for the association between them to produce heightened activation of the target. Priming occurs in ACT* because the prime is still a source of activation when the target appears.

Two lines of evidence are problematic for the Collins and Loftus (1975) model. Ratcliff and McKoon (1981) showed that priming in item recognition was statistically reliable when the stimulus onset asynchrony (SOA) between the prime and the target was as short as 100 ms (no priming occurred at an SOA of 50 ms). This finding suggests that activation spreads very rapidly. In addition, the magnitude of priming at an SOA of 100 ms was the same for prime-target pairs close in network distance and pairs far in network distance. The effects of network distance appeared in the sizes of priming effects at the longer SOAs: More priming eventually occurred for close pairs than for far pairs. In another line of research, Ratcliff and McKoon (1988) showed that the decay of priming could be very rapid, within 500 ms in some circumstances. These findings contradict basic assumptions of the Collins and Loftus (1975) model, but they are quite consistent with Anderson’s (1983) ACT* model.

Compound-Cue Models

Compound-cue models of priming were proposed independently by Ratcliff and McKoon (1988) and by Dosher and Rosedale (1989). The compound-cue model is simply a statement about the contents of retrieval cues. The claim is that the cue to memory contains the target item and elements of the surrounding context. In a lexical decision task, for example, this context could include the prime, or even words occurring before the prime.

The compound-cue model must be combined with a model of memory to make predictions about performance in a task. Models that have figured prominently are the search of associative memory (SAM, Gillund & Shiffrin, 1984), the theory of distributed associative memory (TODAM, Murdock, 1982), and MINERVA 2 (Hintzman, 1986). In all of these models, the familiarity of a cue containing two associated words will be higher than the familiarity of a cue containing two unassociated words. Hence, in a lexical decision task, if the cue contains the target and the prime, familiarity will be higher for a target related to its prime than for a target unrelated to its prime (e.g., lion–tiger vs. tabletiger, respectively). If familiarity is inversely related to response time, basic priming effects can be explained (e.g., Ratcliff & McKoon, 1988).

Distributed Network Models

Relatively recently, several distributed network models of semantic priming have been proposed. These models fall into two broad categories:

In one category of models, which we refer to as proximity models, priming is caused because related primes and targets are closer to each other in a high-dimensional semantic space than are unrelated primes and targets (e.g., Masson, 1995; McRae, de Sa, & Seidenberg, 1997; Moss, Hare, Day, & Tyler, 1994; Plaut & Booth, 2000; Sharkey & Sharkey, 1992). A fundamental assumption in these models is that concepts are represented by patterns of activity over a large number of interconnected units. Related concepts have similar patterns of activity. Semantic priming occurs because in processing a target word the network begins from the pattern created by processing of the prime; this pattern is more similar to the target’s representation when the prime is related than when it is unrelated to the target. In effect, the network gets a head start in processing the target when it is preceded by a related prime. Afew of these models (e.g., Moss et al., 1994; Plaut & Booth, 2000) are able to distinguish semantic priming, which is attributed to overlapping semantic features, from associative priming. Associative priming occurs in these models because the network learns to make efficient transitions from primes to targets that co-occur frequently during training.

The other category of distributed models, which we refer to as learning models, attributes semantic priming to learning that occurs when a word is recognized or is the object of a decision of some kind (e.g., S. Becker, Moscovitch, Behrmann, & Joordens, 1997; Joordens & Becker, 1997). These models also assume that concepts are represented by patterns of activity over a network of units, and that semantically similar concepts have similar patterns of activity. However, in these models semantic priming is caused by incremental learning. Each presentation of a word causes all of the network connections participating in recognition to be altered, so as to increase the probability of producing the same response to the same input. This learning facilitates processing of the word if it reappears, but it also facilitates processing of words with similar representations (e.g., a semantically relatedtarget). Learning decays very slowly and is permanent unless undone by additional learning.This class of models, unlike all other models of priming, predicts that semantic priming should occur over very long lags between presentation of the prime and the target. Data relevant to this prediction are reviewed in a subsequent section of the paper. Proximity may also play a role in these models, especially in explaining priming at short lags.

Major Issues and Findings

Neely (1991) provides the best comprehensive review of research on semantic priming prior to 1991. Our review uses Neely’s as a launching point. We focus on empirical issues and findings that have turned out to be especially important for testing models of semantic priming.

Automatic Versus Strategic Priming

Automatic processes are traditionally defined as those having a quick onset, proceeding without intention or awareness, and producing benefits but not costs. Strategic processes are slower acting, require intention or awareness, and produce both benefits and costs (e.g., Posner & Snyder, 1975).

Semantic priming almost certainly is not caused solely by strategic processes (cf. C. A. Becker, 1980). Semantic priming occurs even when there is only one related prime-target pair in the entire test list (Fischler, 1977a). In addition, at short SOAs, semantic priming occurs between a category name prime and exemplars of that category (e.g., body–leg) even when subjects are told to expect members of a different category (e.g., parts of buildings) to follow the prime (Neely, 1977). Findings such as these are difficult to reconcile with a purely strategic account of priming. Semantic priming, however, is also not purely automatic. Two types of strategic processes have been identified.

Under the appropriate conditions, semantic priming seems to be affected by an expectancy process (e.g., C. A. Becker, 1980; Neely, 1977). Subjects use the prime to generate explicit candidates for the upcoming target or at least expect primes to be followed by semantically related targets. Priming can be amplified because of a speeding up on related trials or a slowing on unrelated trials. Two factors seem to influence the extent to which expectancy processes are used:

1. The SOA between the prime and the target must be sufficiently long to allow expectations to develop. A commonly used index of expectancy is inhibition, or longer response latencies following unrelated primes than neutral primes (e.g., a row of xs, or the words blank or ready). The reasoning is this: An expectancy process will yield an incongruent outcome on unrelated trials because the target is unrelated to the prime. Responses should therefore be slow in the unrelated condition relative to a condition in which expectancies are not generated. A neutral prime condition should provide such a baseline because neutral primes are repeated many times in the test list and are effectively meaningless in the context of the experiment. It is well documented that inhibition is small or nonexistent for SOAs shorter than 300 ms (e.g., de Groot, 1984; den Heyer, Briand, & Smith, 1985; Neely, 1977). In a direct test of expectancy-based priming, Neely (1977) instructed subjects to generate members of a specified category when given a different category name as the prime; for example, subjects were told to generate parts of the body in response to the prime building (and building parts in response to the prime body). Expectancy-based priming occurred at a 700-ms but not at a 250-ms SOA.
2. The second factor that influences expectancy is the relatedness proportion (RP), which is typically defined as the proportion of related trials out of all word prime–word target trials (e.g., Neely, Keefe, & Ross, 1989). At long SOAs, semantic priming and inhibition both increase in magnitude as the proportion of related trials increases; at short SOAs, the effects of RP are reduced or eliminated (e.g., de Groot, 1984; den Heyer, Briand, & Dannenbring, 1983; Tweedy, Lapinski, & Schvaneveldt, 1977). Priming in the naming task also increases with the RP (Keefe & Neely, 1990), suggesting that naming is also influenced by expectancy. It is unknown how low the RP must be to eliminate expectancy. Low values of RPin published studies typically range from .10 to .33.

The second type of strategic process is semantic matching (e.g., de Groot, 1983; Forster, 1981; Neely, 1977; Neely et al., 1989; Seidenberg, Waters, Sanders, & Langer, 1984). Under the appropriate conditions, subjects seem to check for a relation between the target and the prime, responding quickly if such a relation is detected, and slowly if no such relation is detected. In the lexical decision task, the existence of a semantic relation is always informative about the lexical status of the target, as only word targets have related primes.

However, the absence of a relation may or may not be informative depending on the construction of the test list. One measure of the informativeness of the absence of a semantic relation is the nonword ratio (NR), which is the conditional probability that the correct response is nonword given that the (word) prime and the target are unrelated (Neely et al., 1989). As the nonword ratio deviates from .5, the absence of a semantic relation between the prime and the target becomes increasingly informative, signaling a nonword response when it is above .5 and a word response when it is below .5.

The variables that control semantic matching are not well understood. Neely et al. (1989) manipulated the RP and the NR independently in a lexical decision task in which primes were category names and targets were exemplars. The RP was correlated most strongly with priming for typical exemplars (e.g., robin for bird). The NR, however, was correlated with priming for both typical and atypical (e.g., penguin) exemplars, and with nonword facilitation (defined as faster responses to nonwords primed by words than to nonwords primed by a neutral prime). They argued that the effect of RP on priming for typical exemplars was a true expectancy effect, as subjects would be likely to generate typical but not atypical exemplars to category primes. According to Neely et al., the effect of NR was due to semantic matching. The nonword facilitation effects are especially consistent with this interpretation, as, when NR is high, nonword targets will benefit from a bias to respond nonword to targets unrelated to their word primes.

It seems likely that semantic matching is influenced by the RP and the NR. As the RP increases, semantic relations become more noticeable, and as the NR increases, the absence of semantic relations becomes more informative. It is worth pointing out that standard experimental procedures often lead to NRs over .5, as investigators often use equal numbers of word and nonword targets, but only use word primes; hence, the number of word prime–nonword target trials exceeds the number of unrelated word prime–word target trials.

Semantic matching is probably also influenced by the task used. Tasks such as lexical decision that require accumulation of information to make a binary decision are probably more susceptible to semantic matching than are tasks, such as naming, that do not involve an explicit decision (e.g., Seidenberg et al., 1984). McNamara and Altarriba (1988; see also Shelton & Martin, 1992) have argued that semantic matching, as well as expectancy, can be minimized by using a task in which the relations between primes and targets are not apparent to subjects. One method of achieving this goal is to use a sequential or single-presentation lexical decision task. In this task, stimuli are displayed one at a time, and participants respond to each as it appears. Primes precede targets in the test list, but their pairings are not apparent to subjects. Shelton and Martin found that inhibition and backward priming (e.g., prime hop, target bell; discussed later) did not occur in the single-presentation task.

Neely and Keefe (1989) have proposed a three-process hybrid theory of semantic priming that incorporates expectancy, automatic spreading activation, and semantic matching. Not surprisingly,thistheorycanaccountforagreatervarietyofresults than can any one mechanism alone (Neely, 1991). The important contribution of this theory is that it combines a model of automatic, attention-free priming with strategic, attention-laden processes. Viewed in this way, one can see that any of the models of priming outlined earlier in this research paper could be combined with expectancy and semantic matching processes.

In summary, two principal types of strategic processes have been identified, expectancy and semantic matching. Expectancy is minimized at short SOAs and low RPs; semantic matching is minimized with an NR of .5 and, we suspect, low RP as well. Put another way, an investigator interested in the automatic component of priming would be well served by using an SOA less than 300 ms, RP of .20 or less, and NR of .50.

In closing, we should acknowledge that Plaut and Booth (2000) have shown that it may be possible to account for the dependence of inhibition on SOA without invoking an expectancy process. Given all of the evidence implicating the role of strategic processes in semantic priming, it seems likely that any model of priming must incorporate strategic processes of some kind. However, Plaut and Booth’s analysis suggests that a single-mechanism account of priming may be able to explain at least some of the phenomena previously attributed to strategic processing.

Associative Versus Pure Semantic Priming

As noted earlier, the term semantic priming is a catch-all phrase that includes priming caused by many different kinds of relations, including both associative relations and true relations of meaning. Associatively related words are those produced in response to each other in free-association tasks, and they may be semantically related (e.g., dog–cat) or not (e.g., stork–baby). Pure semantically related pairs share semantic features or are members of a common category but are not associatively related (e.g., goose–turkey).

It is well documented that associatively related words prime each other in lexical decision, naming, and similar tasks. The controversial issue has been whether priming occurs in the absence of association. The evidence is mixed. Fischler (1977b) first investigated priming in the absence of association and reported a reliable pure semantic priming effect. However, several subsequent studies (e.g., Lupker, 1984; Moss, Ostrin, Tyler, & Marslen Wilson, 1995; Shelton & Martin, 1992) failed to find pure semantic priming under certain conditions; indeed, Shelton and Martin (1992) concluded that automatic priming was associative, not semantic. Recent experiments by McRae and Boisvert (1998) indicate that previous failures to find pure semantic priming can be attributed to the use of prime-target pairs that were weakly semantically related.

A recent meta-analysis may bring order to this apparent chaos. Lucas (2000) examined the results of 26 studies in which purely semantically related prime-target pairs were used as stimuli in lexical decision or naming (including Stroop) tasks. Most of these studies also included associatively related primes and targets. The average effect size (J. Cohen, 1977), weighted by the number of subjects in each sample, was .25 for pure semantic priming and .49 for associative priming. There was clear evidence therefore that pure semantic priming was present in the studies reviewed and that associative priming was substantially larger than semantic priming. Because associatively related primes and targets were usually related semantically, the larger effect size is best interpreted as an associative boost to priming. Further analyses indicated that the effect size for pure semantic priming was not influenced by the particular type of lexical decision task used, RP, or SOA, suggesting that pure semantic priming was not strategically mediated.

Lucas (2000) also examined whether pure semantic priming varied with type of semantic relation. Category coordinates (e.g., bronze–gold), synonyms, antonyms, and script relations (e.g., theater–play) had similar average effect sizes, ranging from .20 to .27. In contrast, functional relationships (e.g., broom–sweep) had an average effect size of .55. This result supports the hypothesis that functional relations are central to word meaning (e.g., Tyler & Moss, 1997). Perceptually related prime-target pairs, in which primes and targets share referent shape (e.g., pizza–coin), had a very low effect size of .05. This estimate must be treated with caution, however, because only two studies in the corpus examined perceptual priming of this kind.

In summary, although the evidence on pure semantic priming has been mixed, with some studies finding evidence of such priming and others not, Lucas’s (2000) meta-analysis shows that pure semantic priming does occur and, moreover, indicates that it may vary as a function of the type of semantic relation. This conclusion is important because distributed network models of priming strongly predict semantic priming.A subset of these models can also explain associative priming (Moss et al., 1994; Plaut & Booth, 2000). Distributed network models that do not include an associative component will need to be modified to account for the associative boost to priming. Spreading-activation and compound-cue models can easily explain both semantic and associative priming as long as the appropriate relations are represented in memory.

Mediated Versus Direct Priming

Mediated priming involves using primes and targets that are not directly associated or semantically related but instead are related via other words. For example, based on freeassociation norms (e.g., McNamara, 1992b), mane and tiger are not associates of each other, but each is an associate of lion. The associative relation between a prime and a target can be characterized in terms of the number of associative steps or links that separate them: 1-step, or directly related (e.g., tiger–stripes), 2-step (e.g., lion–stripes), 3-step (e.g., mane–stripes), and so on. Models of priming are distinguished based on whether or not they predict priming through mediated relations.

Early experiments suggested that 2-step mediated priming occurred in naming but not in lexical decision (e.g., Balota & Lorch, 1986; de Groot, 1983). Subsequent studies showed that 2-step, and even 3-step, priming could be obtained in lexical decision if the task parameters were selected so as to minimize strategic processing (e.g., McNamara, 1992b; McNamara & Altarriba, 1988; Shelton & Martin, 1992).

Mediated priming is strongly predicted by spreading activation models. Certain versions of compound-cue models can account for 2-step priming, but none predicts 3-step priming (McNamara, 1992a, 1992b). Most distributed network models cannot account for mediated priming of any kind. Possible exceptions are the models proposed by Moss et al. (1994) and by Plaut and Booth (2000). These models learn associative relations between words that co-occur frequently during learning. It is possible that other distributed network models could be augmented with similar mechanisms.

A serious problem exists, however, in interpreting the mediated priming results. Although researchers have made valiant efforts to show that mediated primes and targets are not directly associated and not semantically related (e.g., McNamara, 1992b), there is the nagging possibility that residual associations or semantic relations still exist. This is a big problem because if the primes and targets are directly related in some fashion, all models predict priming between them. The best way to address this issue is in the context of a particular model. For example, McNamara (1992b) showed, using the memory model SAM (Gillund & Shiffrin, 1984), that if direct associations between 3-step primes and targets were high enough to produce priming of the magnitude observed, then these primes and targets would have appeared as mutual associates in a free-association task at a much higher frequency than was observed. This analysis does not prove that the primes and targets were not directly related, and the conclusion is limited to one model of priming (viz., the compound-cue model conjoined with SAM). The contribution exists in demonstrating that a particular model would have difficulty accounting for both the mediated priming and the free-association results. As another example of this approach, Livesay and Burgess (1998) used HAL (discussed in the section on models of semantic memory) to compute semantic distances between the 2-step mediated primes and targets developed by Balota and Lorch (1986) and subsequently used by McNamara and Altarriba (1988). Average semantic distance was higher between mediated primes and targets than between unrelated primes and targets. In addition, they found no relation between the magnitude of mediated priming and lexical co-occurrence frequency, contradicting predictions of McKoon and Ratcliff (1992). These results lead us to conclude that mediated priming remains a challenge to many models of semantic priming.

Effects of Lag

Lag refers to the number of items that intervene between the prime and the target. The standard priming paradigm uses a lag of zero; the target immediately follows the prime. Many studies have examined priming at lags of one, two, and even greater. The early literature on lag effects was ambiguous (e.g., Masson, 1991). Subsequent investigations indicated that priming occurred across a lag of one but not two (e.g., Joordens & Besner, 1992; McNamara, 1992b), although Masson (1995) did not obtain lag-1 priming in naming.

Recent experiments indicate that semantic priming may occur over lags much greater than one or two items. For theoretical reasons (discussed in the section on models of semantic priming), S. Becker and Joordens (S. Becker et al., 1997; Joordens & Becker, 1997) hypothesized that semantic priming could be obtained at long lags if the primes and the targets were strongly semantically related and the task engaged semantic processing to a high degree. They constructed prime-target pairs that were semantically similar (e.g., pontoon–raft, tulip–rose) and used several methods to increase the semantic processing of target words. S. Becker et al. (1997) used an animacy decision task in which participants were required to decide whether each word referred to a living or a nonliving entity; Joordens and Becker (1997) used a lexical decision task in which nonwords were very word-like (e.g., brane). Semantic priming was obtained in these experiments at lags of 4, 8, and even as high as 21.5.

Priming at long lags is predicted by learning models, but it is a serious problem for all other models of priming. In principle, spreading activation models could explain such priming by making the decay of activation very slow, but this assumption would be inconsistent with other findings suggesting that activation decays quickly. Moreover, slow decay would probably leave so much residual activation in memory that basic semantic priming effects could no longer be predicted. Compound-cue models would need cues of between 23 and 24 items to explain priming at a lag of 21.5 (prime + intervening items + target). Cues of this size strain credibility. Proximity models explain priming across intervening items by assuming that the semantic pattern of the prime is not completely replaced by semantic patterns of intervening items (e.g., Masson, 1995; Plaut & Booth, 2000). This mechanism almost certainly will not work with lags greater than one or two items.

There are several reasons to question these findings, however. First, the results are unstable. Joordens and Becker (1997) obtained lag-8 priming in two experiments but did not obtain it in another two experiments. Second, and more important, the priming observed in these studies has peculiar properties. Priming either did not decay with lag or decayed rapidly with lag, and yet priming at the shortest lag did not differ in these situations. For example, in their second experiment, Joordens and Becker varied lag over the values 0, 1, 2, 4, and 8. One condition was designed to produce long-term semantic priming and used a lexical decision task with difficult nonwords (e.g., brane). This condition yielded 45 ms of priming at lag 0, and there was no evidence of decay across lags; for example, priming was 41 ms at lag 8. Another condition was designed not to produce long-term semantic priming and used a lexical decision task with easier nonwords (e.g., brene). This condition yielded 27 ms of priming at lag 0, which quickly decayed to nonsignificant levels. The 18-ms difference in priming at lag 0 for these two conditions did not approach statistical significance. This pattern of results is difficult to explain even in the learning model. Why should priming of comparable initial magnitude decay slowly in one case but quickly in another?

Joordens and Becker (1997) proposed several explanations of these results that relied on dual mechanisms, but none was compelling. Their preferred model incorporated a quickly decaying associative priming mechanism with a long-term learning mechanism. Priming in the easy nonword condition would be attributed to the associative mechanism alone, whereas priming in the difficult nonword condition would be attributed to the combined effects of both mechanisms. Even this model, however, is not consistent with their findings, as it still predicts some decline in priming with lag: At lag 0, both associative and learning priming would occur, whereas at longer lags, only learning priming would occur.

In summary, if priming at long lags holds up under additional experimental scrutiny, and if the paradoxical results obtained by Joordens, Becker, and their colleagues can be explained, long-term priming will provide compelling evidence in support of distributed-network learning models and virtually insurmountable evidence against other models of semantic priming.

Forward Versus Backward Priming

Associations between primes and targets can be asymmetric. Backward priming refers to the situation in which the association from prime to target is weak, but the association from target to prime is strong (e.g., baby–stork). Koriat (1981) was the first to investigate backward priming, and he obtained equal amounts of priming in the forward (e.g., stork–baby) and the backward (e.g., baby–stork) directions. This result is surprising, because if priming depends on strength of association, it should be larger in the forward than in the backward direction. A perusal of the backward priming literature reveals the following observations and findings.

One of the difficulties in comparing results across studies is that different materials have been used. Several studies have used asymmetrically associated, semantically related primes and targets (e.g., lamp–light, apple–fruit); other studies have used semantically unrelated compound words (e.g., fruit–fly, sand–box); and still others have used a mixture of these types of stimuli. Given the findings reviewed earlier on pure semantic priming, one would expect to find some priming in the forward and in the backward directions for semantically related pairs, regardless of differences in associative strength. In contrast, one wonders why priming would occur at all for semantically unrelated compounds, unless it is strategically mediated. Of the 20 compounds introduced into the literature by Seidenberg et al. (1984), and subsequently used by Shelton and Martin (1992) and by Thompson-Schill, Kurtz, and Gabrieli (1998), 18 prime words appear in the Nelson, McEvoy, and Schreiber (1991) free-association norms. The associative strength in the forward direction (e.g., fruit–fly) has a modal value of 0 and a mean of .02! These items are therefore neither semantically related nor associated.

In fact, there is good evidence that forward and backward priming for compounds is produced by strategic processes. Priming does not occur in either direction for compounds when conditions are consistent with automatic priming (Shelton & Martin, 1992; Thompson-Schill et al., 1998). In strategic conditions, processing of forward and backward associations is correlated with physiological indices of strategic processing (Chwilla, Hagoort, & Brown, 1998). Those studies reporting reliable priming for compounds in the forward or the backward directions (Kahan, Neely, & Forsythe, 1999; Seidenberg et al., 1984; Shelton & Martin, 1992) employed task parameters in the strategic regime (e.g., high RP, high NR, or long SOA). We include in this mix experiments using the naming task, as there is evidence that naming is not immune to strategic processing (e.g., Keefe & Neely, 1990).

Asymmetrically associated, semantically related pairs (e.g., lamp–light) seem to prime each other in both directions, and there is weak evidence that priming is greater in the forward than in the backward direction. Only three published studies have examined priming in both directions for semantically related pairs (Chwilla et al., 1998; Koriat, 1981; Thompson-Schill et al., 1998), and only one of them has used procedures consistent with automatic priming (ThompsonSchill et al.). Across all three studies, priming was 40% larger, on the average, in the forward than in the backward direction; in the experiments by Thompson-Schill et al., the difference was approximately 30%.

Finally, there is evidence that backward priming in the naming task may depend on the SOA (Kahan et al., 1999; Peterson & Simpson, 1989). For example, Kahan et al. obtained backward priming at an SOA of 150 ms but not at an SOAof 500 ms. These findings are difficult to interpret, however, because the experiments have used RPs of at least .5. In fact, only one experiment (Thompson-Schill et al., 1998) has examined backward priming in naming under near automatic conditions, and it only used one SOA(200 ms). Hence, the apparent dependence of backward priming in naming on SOA may be produced by strategic processing in this task.

In summary, semantic priming does not seem to occur for compounds unless the conditions are ripe for strategic processing, whereas priming occurs in both directions for asymmetrically associated, semantically related pairs, and there is some evidence that the magnitude of priming tracks associative strength. Because only the latter results seem to be caused by automatic processes, only they are crucial for testing models of priming.

Spreading activation models can account for these results as long as appropriate semantic and associative relations exist in memory (e.g., symmetric semantic relations but asymmetric associative strengths). The predictions of compound-cue models depend on which model of memory serves as the base. The two models that have figured most prominently in investigations of priming are SAM (Gillund & Shiffrin, 1984) and TODAM (Murdock, 1982). SAM cannot predict asymmetric priming unless the primes and the targets differ in the strength of the association between the words as cues and their representations in memory. TODAM also has difficulty explaining asymmetric priming because associations are modeled by a commutative operation, convolution. Hence, a demonstration of reliable asymmetric priming with primes and targets of equal word frequency would be problematic for these models. All of the distributed network models predict priming for semantically related primes and targets, but only two (Moss et al., 1994; Plaut & Booth, 2000) have an associative mechanism that would allow them to predict greater priming in the forward than in the backward associative direction.

Subliminal Priming

Several researchers have reported evidence that semantic priming occurs even when the prime is presented under conditions in which it cannot be identified or its presence cannot be detected (e.g., Marcel, 1983). After conducting a comprehensive review of this literature, Holender (1986) concluded that the effects were unreliable and that the stimuli had probably been consciously identified (also see Cheesman & Merikle, 1984). More recent studies addressed many of these problems, but effects were still small and inconsistent (e.g., Greenwald, Klinger, & Schuh, 1995).

Greenwald and his colleagues (e.g., Draine & Greenwald, 1998; Greenwald, Draine, & Abrams, 1996) have recently claimed that robust unconscious priming effects can be obtained under the proper experimental conditions. An important feature of these experiments is that they used evaluative or gender judgments as the priming task rather than standard semantic priming tasks (e.g., lexical decision or naming). For example, in the evaluative judgment task, participants judged whether words had positive or negative meanings (e.g., happy vs. vomit). Priming was assessed by examining the effect of the prime’s category membership (e.g., positive vs. negative) on responses to targets. Greenwald and his colleagues found that under appropriate conditions primes increased the probability of responding in a manner consistent with their category membership even when direct perception of the primes approached zero sensitivity. They attributed this result to the unconscious activation of the meaning of the prime.

There are at least two reasons to question this conclusion, however. First, Klinger, Burton, and Pitts (2000) replicated the priming effects obtained by Greenwald and his colleagues (e.g., Draine & Greenwald, 1998; Greenwald et al., 1996) but also showed that semantic priming of the lion–tiger variety did not occur in the same paradigm. Second, Abrams and Greenwald (2000) have shown that the priming obtained in the basic paradigm does not occur unless primes previously occur as targets; the effect may be due to procedural learning in the task. Our (admittedly conservative) conclusion is that there is little or no convincing evidence that the meaning of a word can be activated unconsciously.

Prime Task Effects

Given that priming occurs when participants read the prime but make no response to it, one might predict that semantic priming would occur regardless of the task performed on the prime. In fact, this is not true. M. C. Smith, Theodor, and Franklin (1983) showed that semantic priming was eliminated if participants searched the prime for a letter or responded whether or not an asterisk was next to the prime (also see Friedrich, Henik, & Tzelgov, 1991; Henik, Friedrich, & Kellogg, 1983; Henik, Friedrich, Tzelgov, & Tramer, 1994; M. C. Smith, 1979). A general conclusion from these studies is that if attention is directed away from the semantic level early in the processing of the prime, semantic priming is eliminated or attenuated (e.g., Stolz & Besner, 1996). In a related line of research, Besner and Stolz (1999) demonstrated that Stroop interference was reduced in magnitude if attention was directed to individual letters of a word (rather than to the whole word).

These and related findings led Stolz and Besner (1999) to conclude that attentional control is needed to activate the meanings of words. They argue that attention determines how activation is distributed across levels of representation (e.g., letter, word, semantic) during word recognition. The attentional mechanisms implied by this explanation are qualitatively different from those implied by the traditional distinction between automatic and strategic priming (as discussed in a previous section); in particular, they must be fast acting and need not be conscious. Prime-task effects create difficulties for all of the models of priming. The fundamental problems are that the models do not cast the proper roles for attention, or they do not distinguish between levels of representation in a manner that would allow, for instance, attention to be directed to one level (e.g., letter) but not to another (e.g., semantic), or both. These problems are not insurmountable, but they are not trivial to solve either.

Neely and Kahan (2001) have recently argued that primetask effects may be caused, at least in part, by effects of spatial attention on visual feature integration. The hypothesis is that when attention is directed to individual components of prime words, such as letters, the visual features of unattended letters may not be properly integrated, and hence the primes may not be perceptually encoded as words. Semantic activation of the primes would not be expected under such circumstances. If Neely and Kahan’s hypothesis is correct, then prime-task effects are not so problematic for the models of priming.

Global Context Effects

Recent experiments indicate that the global context established by discourse or by the types of semantic relations appearing in a test list can affect semantic priming. For example, McKoon and Ratcliff (1995) placed a small number of prime-target pairs related in a particular way (e.g., opposites, close–far) in a list in which over half of all prime-target pairs were related in a different manner (e.g., synonyms, mountainhill). Semantic priming in lexical decision and in naming was virtually eliminated for the mismatching items. Hess, Foss, and Carroll (1995) obtained similar results by varying the global context established by short vignettes preceding target words.

Although one might be tempted to conclude from these findings that all contextual facilitation is determined by global context (e.g., Hess et al., 1995), this conclusion is not justified. There is just too much evidence that semantic priming occurs between strongly related words in the most infelicitous of conditions (e.g., Fischler, 1977a; Neely, 1977). We suspect that appropriate follow-up studies will show effects of local context in addition to global context. The contribution of these studies is to demonstrate that semantic priming is modulated by relations external to the word pairs. These global context effects are a serious challenge to all existing models of semantic priming.

Summary

Given that we used Neely’s (1991) review as a starting point, it is appropriate to ask what has been learned about semantic priming since the publication of that chapter.

First, a new class of models of priming, namely, distributed network models, has been developed. One member of this class of models, the learning models, can explain what may turn out to be the most important new finding on semantic priming, namely, semantic priming over very long lags.

Second, a great deal more has been learned about several important priming phenomena: (a) There is probably a better understanding of the conditions that contribute to automatic versus strategic priming. (b) It is now clear that pure semantic priming occurs, and there is evidence that it is produced by automatic processes. (c) Mediated priming has now been replicated by several investigators using a variety of tasks and procedures. (d) Priming across lags of unrelated intervening items has been replicated in several studies. Moreover, there is new evidence that semantic priming can occur over very long lags. (e) Several new investigations of back ward priming have appeared, and the results suggest that backward priming for compounds (e.g., hop–bell) is produced by strategic processes, whereas backward priming between asymmetrically associated semantically related words (e.g., light–lamp) is caused by semantic overlap. (f) An entirely new line of research on subliminal priming has appeared, and it seems to converge on the conclusion that semantic priming does not occur unconsciously. (g) There is a better understanding of the role of attention in semantic priming. Finally, (h) a new line of research indicates that semantic priming is affected by the global context in which prime-target pairs appear.

Third, although none of the models can account for all of the major priming results, there are reasons to be optimistic about future model development. Assuming, for the moment, that long-term semantic priming turns out to be a robust phenomenon, then distributed-network learning models offer an appealing foundation for model development. If these models can be augmented with associative priming mechanisms and appropriate attentional processes, they will go a long way toward explaining the major findings in the literature.

Repetition Priming

Whereas semantic priming refers to a facilitation in performance between different items on the basis of shared meaning or association, repetition priming refers to facilitated performance based on a previous encounter with the same stimulus. Essentially, repetition priming reflects the degree to which a single exposure to a stimulus during a study session leads to faster or more accurate processing of that stimulus at a later test (Tulving & Schacter, 1990).

Research on repetition priming has developed largely from studies involving patients with anterograde amnesia. This neurological disorder (or collection of disorders) is characterized by a severely impaired ability to form new explicit memories (for review, see Squire, 1987). This type of amnesia typically accompanies damage to the medial temporal lobes (e.g., for review, see Squire, 1992) or to the diencephalic midline structures (as in Korsakoff’s syndrome; for review, see Oscar-Berman, 1984; Shimamura, 1989). Amnesic patients show an impairment of the ability to explicitly recall events that occur after the onset of their amnesia, despite intact intellectual, language, and social skills. In spite of showing severely degraded performance on tests of explicit memory, which require conscious recollection, such as free recall, cued recall, recognition, and paired-associate learning, amnesic patients exhibit intact performance on measures of repetition priming, such as word-stem completion, picturefragment completion, and picture naming (e.g., Cave & Squire, 1992; N. J. Cohen & Squire, 1980; Graf, Squire, & Mandler, 1984; Shimamura, 1986, 1993; Squire, 1987; Warrington & Weiskrantz, 1968).

In a now-classic study examining repetition priming in amnesic patients, Graf et al. (1984) used a word-stem completion paradigm, in which participants see partially completed words (e.g., ele_____) at test and are asked to fill in the blanks to form the first word that comes to mind. Although no reference is made to an earlier list of words, participants are more likely to complete the string elephant if they saw elephant during an earlier study session than if they did not. Graf et al. found that normal and amnesic participants showed equal levels of repetition priming on this task. Explicit memory instructions, however, changed the pattern of results. When asked to complete the word stems with study list words, normal participants improved dramatically in their ability to produce study words, whereas amnesic participants did not improve at all. This study and others have shown that robust repetition priming can occur in the absence of explicit memory (for a review, see Shimamura, 1986), and have not only identified areas of the brain crucial to explicit memory but also contributed to the notion that repetition priming may be a form of implicit memory subserved by regions of the brain other than those damaged in amnesia (Cave & Squire, 1992; Schacter, 1990; Squire, 1992; Squire et al., 1993). However, as shall be seen, the notion that repetition priming represents a distinct memory system has been at the center of much controversy. Indeed, not all researchers even acknowledge that repetition priming reflects an aspect of memory.

Models of Repetition Priming

Theories of implicit memory have typically not been concerned with specific processing assumptions, and few research studies have attempted to provide detailed descriptions of the processes underlying repetition priming. As a consequence, model development is not as advanced as in other areas of memory research.

Logogen Model

According to Morton’s (1969) model of word recognition, words are mentally represented by feature counters, called logogens. An incoming word stimulus causes information to accumulate in the counters for all words that share properties with that stimulus. A word is recognized when the amount of information accumulated in a logogen exceeds the threshold value for that logogen. Repetition priming can be explained as the lowering of the threshold for a previously encountered word (or, equivalently, as the raising of the logogen’s resting activation level).

Counter Model

The counter model (Ratcliff & McKoon, 1997) is a variant of the logogen model and was developed to explain repetition priming in perceptual identification. According to the model, each word is represented by a counter, and a decision is made based upon the accumulation of counts. Counts can correspond to perceptual features or to noise (null counts). Null counts are needed to allow the system to respond when there is little or no perceptual information coming from the stimulus. The characteristic of the model that allows it to explain repetition priming is that counters can become attractors of counts. The counter for a previously studied word can steal counts from the counters of similar words. This mechanism produces a pattern of bias in repetition priming because theft of counts is based on similarity and occurs regardless of whether or not the repeated word is the target. Consider, for example, forced-choice perceptual identification in which a target word (e.g., lied) is briefly flashed and the subject must then choose between two similar options (e.g., lied vs. died). If the flashed target is lied, prior study of lied causes an increase in performance; but if the flashed target is died, prior study of lied causes a decrement in performance (because its counter steals counts from died’s counter). Put another way, people are biased to see the word that was studied previously, even when it is not the target. A potential limitation of the counter model is that it only applies to perceptual identification, which is just one of many tasks in which repetition priming is observed. Of course, this need not present a problem if one takes the view that repetition priming is merely the by-product of the task in which the effect appears (Ratcliff & McKoon, 1997). In this view, repetition priming in perceptual identification may be caused by entirely different mechanisms from those responsible for repetition priming in another task, such as fragment completion. Another problem for the counter model is that there is evidence that prior study can produce increased sensitivity in addition to bias (e.g., Bowers, 1999; Wagenmakers, Zeelenberg, & Raaijmakers, 2000). Although the counter model can be modified to account for these findings (Ratcliff & McKoon, 2000), this change represents a major conceptual shift in the model.

Instance Theory

Logan (1988, 1990) has taken a different approach to elucidating the processes that may underlie repetition priming. In an effort to bridge the gap between research on repetition priming (the effects of one prior exposure on performance) and research on automaticity (the effects of very many exposures on performance), Logan proposed that repetition priming may be a form of skill acquisition governed by a power function of the number of practice trials. Essentially, Logan has suggested that repetition priming and automaticity may reflect two ends of the same continuum.According to Logan’s instance theory, initial performance on a task is determined by a general problem-solving algorithm. As the task progresses, every encounter with a stimulus is stored as a separate instance, even if it is identical to a previous episode. Eventually a level of proficiency with the task may be reached at which the algorithm can be abandoned and responses can be made solely on the basis of instances (i.e., automatically). Presumably, the retrieval of instances can be more efficient than performance of the algorithm. Performance between these two extremes may be automatic for some trials, but not for others. As more instances of a particular stimulus are encoded, the likelihood that the stimulus will receive an automatic response increases. Thus, repetition priming reflects the increased likelihood of an automatic response’s following a single prior exposure to a stimulus.Aproblem for the instance theory is that experiments by Kirsner and Speelman (1996) have provided evidence suggesting that repetition priming can be indifferent to practice and may in fact be a one-shot effect. These findings certainly cast some doubt on the notion that repetition priming and skill acquisition reflect the operation of the same underlying mechanism.

Distributed Network Models

Repetition priming can be explained in distributed network models in much the same way as semantic priming is explained in these models. Indeed, repetition priming can be viewed as an example of semantic priming in which the prime’s and the target’s semantic representations (as well as orthographic and phonological representations) are identical. Relatively few distributed network models have been applied to repetition priming (but see McClelland & Rumelhart, 1985; Stark & McClelland, 2000), although repetitionpriming effects are often interpreted in the context of these models (e.g., Rueckl, Mikolinski, Raveh, Miner, & Mars, 1997). Distributed network models of repetition priming have not been investigated in as much depth as have distributed network models of semantic priming, and little is known about their abilities to account for the major results in the literature.

Major Issues and Findings

As mentioned earlier, research on repetition priming began with studies of patients suffering from impairments of explicit memory. It is not surprising, therefore, that from these earliest observations repetition priming has often been viewed as a form of memory, a type of implicit memory that remains intact in amnesics. Today the term implicit memory encompasses a variety of phenomena (e.g., semantic priming, classical conditioning) whose common feature is the influence of prior episodes on behavior without effortful, or explicit, retrieval of those episodes (for reviews, see Richardson-Klavehn & Bjork, 1988; Squire et al., 1993). Repetition priming represents one of the most thoroughly researched of these phenomena. The discussion that follows highlights some of the major findings in the repetition-priming literature, as well as the primary theoretical approaches that have guided the research.

Dissociations from Explicit Memory

If studies of implicit and explicit memory measures produced differences only in an amnesic population, the results of such studies might be of limited interest. However, similar dissociations between implicit and explicit memory performance have been repeatedly demonstrated in normal participants (for a review, see Richardson-Klavehn & Bjork, 1988). Dissociations between implicit and explicit memory performance in normal participants, however, are typically of a different nature. Rather than demonstrating a presence of implicit memory and an absence of explicit memory as in amnesic patients, dissociations in normal participants are typically demonstrated through differential effects of manipulating an independent variable on measures of implicit and explicit memory.

Levels of Processing. Jacoby and Dallas (1981) demonstrated that manipulating the level of processing of study words did not affect the magnitude of repetition priming in a perceptual identification task (i.e., accuracy of correctly identifying briefly presented stimuli), but it produced large effects on an explicit recognition task (prior semantic processing of words yielded better performance than phonemic or orthographic processing). Insensitivity of repetition priming to level of processing manipulations has also been demonstrated using word fragment completion (e.g., Graf & Mandler, 1984), lexical decision (e.g., Monsell, 1985), perceptual identification of pictures (e.g., Carroll, Byrne, & Kirsner, 1985) and picture naming (e.g., Carroll et al., 1985).

Effects of Delay. Measures of implicit and explicit memory also show differential effects of delay. Several studies manipulating retention interval have found that repetition priming is less affected by delay than explicit memory. For instance, Jacoby and Dallas (1981), Tulving, Schacter, and Stark (1982), and Mitchell and Brown (1988) found that repetition priming persisted with little change across delays of days and weeks (using perceptual identification, word fragment completion, and picture naming tasks, respectively), whereas recognition performance declined sharply across the same delays. Researchers have since demonstrated intact repetition priming for pictures at delays up to 48 weeks (Cave, 1997) and for words at delays up to 16 months (Sloman, Hayman, Ohta, Law, & Tulving, 1988). These findings suggest that repetition priming can be a relatively permanent form of long-term memory rather than a temporary facilitation due to a recent encounter with a stimulus.

Developmental Differences. Research on developmental differences between implicit and explicit memory has been another source of observed dissociations. Many studies have been performed examining developmental dissociations between repetition priming and explicit memory in populations as young as 3 years (e.g., Drummey & Newcombe, 1995; Greenbaum & Graf, 1989), 4 years (e.g., Hayes & Hennessy, 1996), and 5 years (e.g., Carroll et al., 1985). In studies comparing repetition-priming performance of children and adults, equivalent levels of repetition priming were detected across all tested age levels. In contrast, explicit memory performance continued to show developmental improvements up to at least age 12 (e.g., Carroll et al., 1985). Similarly, different developmental trends between explicit and implicit memory have been detected in the aged, with elderly participants typically showing decreases in explicit memory performance relative to younger adults, despite showing equivalent levels of repetition-priming performance (e.g., Graf & Ryan, 1990; Mitchell, 1989); for a discussion of specific implicit memory impairments in elderly participants with Alzheimer’s disease, see Gabrieli et al. (1999).

Multiple Systems Versus Processing Theories

Multiple Systems Theories. One way to account for dissociations between implicit and explicit memory has been to postulate separate memory systems in the brain for different types of memory (e.g., N. J. Cohen & Squire, 1980; Schacter, 1990; Squire, 1986; Tulving & Schacter, 1990). Researchers postulating distinct memory systems derive support for this notion in large part from studies with amnesic patients. According to this view, the brain damage in amnesia selectively affects the memory system for conscious recollection, leaving the system (or systems) responsible for other forms of memory intact. Evidence from studies involving amnesic patients suggests that different neural structures underlie performance on tests that rely on different kinds of memory. Because these memory systems operate largely independently of each other, dissociations between performance on tasks that utilize different systems are to be expected. For example, the hippocampus seems to play a crucial role in explicit memory, yet implicit memory performance seems to be unaffected by damage to the hippocampus. The hippocampus, then, must be part of an explicit memory system. Thus, dissociations between different measures of memory are explained by appealing to different memory systems. Based on this view, a taxonomy of memory can be established to classify measures based on the neural mechanisms with which they are associated.

It is important to note, however, that a single (or one-way) dissociation between memory phenomena, such as that observed in amnesics, does not necessarily imply separate memory systems. The data from studies with amnesics could be explained within a single-system framework, for instance, by arguing that explicit memory tasks are more demanding of the neurological resources of a single memory system than implicit tasks. Thus, damage to the memory system, as occurs with amnesia, may leave the system too injured to meet the demands of an explicit memory task, yet not so injured as to affect performance on a less demanding implicit memory task. Many such functional hierarchies might be imagined that incorporate implicit and explicit memory into a single system. Any such structure, however, can only predict the one-way dissociation of intact implicit memory with impaired explicit memory. In the previous example, for instance, implicit memory is more resilient because it is less demanding on the memory system. Thus, if damage to the system were to occur such that implicit memory were impaired, explicit memory would necessarily be impaired because, according to this formulation, the demands on the system are greater for explicit memory tasks than for implicit memory tasks. However, evidence for a double dissociation between implicit and explicit memory has been reported by Gabrieli and colleagues (Gabrieli, Fleischman, Keane, Reminger, & Morrell, 1995; Keane, Gabrieli, Mapstone, Johnson, & Corkin, 1995), who have studied patients with occipital lobe lesions. These patients demonstrated impaired repetition-priming performance in a perceptual identification task despite intact explicit memory performance. The results of these studies have been used to support the notion that implicit memory phenomena such as repetition priming are mediated by brain systems separate from those mediating explicit memory. The results of the experiments by Gabrieli and colleagues provide strong evidence that the processes supporting repetition priming do not necessarily contribute to explicit memory performance. Thus, repetition-priming effects should not be interpreted in terms of degraded explicit memory performance. In addition, it should be noted that these patients’ ability to perform a perceptual identification task without eliciting repetition priming presents difficulties for models such as the counter model, which assumes that repetition priming occurs as a by-product of performing this task.

On the basis of observed functional and stochastic dissociations between implicit and explicit memory, as well as the evidence from amnesics and other patients with brain damage, theorists have proposed a multiple systems view of memory (e.g., Schacter, 1992; Squire, 1992; Tulving & Schacter, 1990), which holds that neurologically distinct systems subserve different types of memory. The patterns of data have suggested to many that implicit memory is supported by systems distinct from those required for the formation of explicit memories.

The multiple systems view also divides implicit memory into subsystems. Perhaps the best elaborated multiple systems account is that of Tulving and Schacter (1990), who propose a set of neurologically distinct perceptual representation systems (PRS), each of which is designed to encode a particular type of information. Each PRS is presemantic— encoding perceptual information without the necessity for the stimulus to be processed semantically—and supports repetition priming on tasks that use that information. Schacter and his colleagues postulate at least three such systems: a system that encodes information about object parts and their relations in the form of structural descriptions (e.g., Humphreys & Quinlan, 1987; Riddoch & Humphreys, 1987) and supports repetition priming for objects (e.g., Schacter, Cooper, & Delaney, 1990a, 1990b); a visual word-form system that encodes graphemic word information and supports repetition priming for visually presented words (e.g., Marsolek, Kosslyn, & Squire, 1992); and a similar auditory word-form system (e.g., Church & Schacter, 1994). Other systems presumably support repetition priming in more conceptual tasks, although the focus of the multiple systems view has thus far been on perceptual tasks. However, whether implicit and explicit memory are subserved by separate systems at all is a heavily debated issue (cf. Blaxton, 1989; Roediger, 1990; Shimamura, 1990).

Processing Theories. Many researchers have chosen to distinguish memory phenomena on the basis of the different cognitive processes required by the memory tests (e.g., Graf & Mandler, 1984; Roediger & Blaxton, 1987; Roediger, Weldon, & Challis, 1989). Rather than assuming that implicit and explicit tests access separate memory systems, processing theories assume that memory tests are composed of various component processes, and dissociations between performance on memory tests reflect the operation of different processes.

Perhaps the most commonly stated processing account of memory is embodied in the principle of transfer appropriate processing (TAP; e.g., Morris, Bransford, & Franks, 1977; Franks, Bilbrey, Lien, & McNamara, 2000). A primary assumption of TAP is that performance on a memory test benefits to the extent that the cognitive operations at test overlap with those engaged during initial learning. In general, dissociations between performance on explicit and implicit memory tests are characterized in terms of a distinction between conceptually driven processes and data-driven processes (Roediger et al., 1989). Explicit memory tasks typically (but not always) depend on conceptual processing that is assumed to be sensitive to delay and depth of processing manipulations, whereas implicit memory tests usually depend on datadriven processing that is assumed to be insensitive to these factors. Along the same lines as TAP, Jacoby (1991) has proposed that implicit memory involves automatic processes, whereas explicit memory requires consciously controlled processes. Although Jacoby argues against equating specific memory tests with proposed cognitive processes, he suggests the possibility that special populations, such as amnesics, may show a deficit in intentional processing but preserve automatic or unconscious forms of memory. Jacoby argues that both automatic and controlled processes are always operating, and he has postulated a framework called process dissociation that is designed to parse out the relative contributions of each process to performance on a given task.

There has been much debate in recent years concerning the issue of whether memory should be characterized in terms of memory systems or in terms of cognitive processes (e.g., Graf & Ryan, 1990; Mitchell, 1993; Roediger, 1990; Roediger et al., 1989). It is not uncommon for multiple memory systems and processing views to be represented in the implicit memory literature as rival hypotheses. Several researchers, however, have pointed out that these two perspectives are not necessarily incompatible (Schacter, 1990; Shimamura, 1989, 1993). Shimamura (1993) argues that the debate between multiple-systems and TAP views appears to be the result of scientists’ working from two different perspectives. Processing views, such as TAP, for instance, are typically championed by researchers in cognitive psychology, whereas multiple-systems views are often forwarded by researchers in neuroscience. The argument for a processing view—in contrast to a multiple-systems view—is often simply a matter of emphasis. Shimamura points out that a processing theory becomes a systems view when it attempts to identify the neural circuitry associated with a process. Likewise, a systems view becomes a processing view when it attempts to identify the process that is subserved by some neural circuitry.

Thus, it may make little theoretical difference whether memory is characterized in terms of multiple brain systems or cognitive processes. Ratcliff and McKoon (1996), however, have pointed out that focusing on multiple memory systems has moved memory to the foreground, and put into the background an understanding of the mechanisms that mediate memory performance. Likewise, the debate between systems and processing views of memory has moved to the foreground research focused on supporting one view over the other, rather than using the broader perspective afforded by both views taken together to forward our understanding of memory phenomena.

Conceptual Versus Perceptual Priming

Primarily on the basis of the processing demands of a repetition-priming task, researchers have made a distinction between conceptual and perceptual repetition priming (e.g., Roediger & Blaxton, 1987; Roediger et al., 1989). Tasks that involve analysis of stimulus meaning engage conceptual processes, and tasks that involve analysis of stimulus form engage perceptual processes.

Conceptual repetition priming is largely unaffected by changes in the perceptual qualities of a stimulus between study and test, and it is greater following conceptual elaboration at study, such as encoding the meaning of study items. Test tasks that have been used to measure conceptual repetition priming include category exemplar generation and answering general knowledge trivia questions (e.g., Blaxton, 1989; Rappold & Hashtroudi, 1991). Although conceptual processes are thought to also underlie performance on most explicit memory tasks, dissociations between implicitconceptual and explicit-conceptual tasks have been reported in normal and brain-damaged participants (e.g., Graf, Shimamura, & Squire, 1985; McDermott & Roediger, 1994). Although these results have been taken to suggest that separable processes underlie conceptual implicit and explicit memory, relatively little is known about the processes underlying conceptual priming (Vaidya et al., 1997).

Although the majority of research on perceptual repetition priming has been in the visual domain, repetition priming has also been examined in the auditory domain. For instance, Schacter, Church, and their colleagues (Church & Schacter, 1994; Schacter, Church, & Treadwell, 1994; Schacter, Church, & Bolton, 1995) undertook a systematic investigation of repetition priming across changes in a variety of auditory dimensions. Their participants listened to lists of words recorded from a single speaker. At test, they attempted to identify old and new words embedded in noise. Repetition priming in this paradigm is evidenced by improved accuracy for old words (Jackson & Morton, 1984). They found that repetition priming was reduced (but not eliminated) by changes in speaker, emotional or phrasal intonation, and fundamental frequency, but not by changes in volume (Church & Schacter, 1994). Repetition priming was attributed to the operation of an auditory word-form system specialized to encode frequency information. Whether auditory repetition priming is consistently sensitive to frequency information is still uncertain; the effects of speaker are not observed in amnesic patients (Schacter et al., 1995) and are sometimes not seen in normal participants (Jackson & Morton, 1984).

Perceptual Specificity in Repetition Priming

As mentioned previously, the majority of repetition priming research has focused on perceptual repetition priming in the visual modality. Common tests for visual perceptual repetition priming include word-stem completion, fragment completion (word and picture), lexical decision, perceptual identification (word and picture), and picture naming. Generally, perceptual repetition priming at test is unaffected by conceptual elaboration of study items (e.g., shallow vs. deep processing) but is reduced when perceptual characteristics are changed between study and test. For instance, changes in pictorial exemplar (e.g., from jet to biplane) can reduce perceptual repetition priming, as can changes in symbolic form (e.g., from picture to word) or presentation modality (e.g., from auditory to visual; e.g., Biederman & Cooper, 1991b; Blaxton, 1989; Weldon, 1991).

Interestingly, Easton, Srinivas, and Greene (1997) demonstrated robust repetition priming for words between visual and haptic modalities (words were printed in raised characters that were felt-like to the touch). Easton et al. speculated that vision and haptics may both be adapted for spatial or object discrimination and may share many of the same processing demands and representational characteristics. Specifically, vision and haptics may share geometric representations, unlike the phonological representations in audition. Thus, repetition priming can occur across modalities if the modalities share common representations (or representational characteristics), but repetition priming is attenuated if the representations of a stimulus in different modalities are also different.

Despite the strong evidence for perceptual specificity in priming, some studies have indicated that repetition priming involving pictorial stimuli may be unaffected by a broadrange of perceptual manipulations such as changes in size, location, direction of face, color, and illumination (e.g., Biederman & Cooper, 1991a, 1991b, 1992; Cave, Bost, & Cobb, 1996; Cooper, Schacter, Ballesteros, & Moore, 1992; Srinivas, 1996a). Some of these findings have been interpreted as evidence that object identification is not sensitive to certain stimulus attributes, and therefore priming is not affected by these attributes. However, Srinivas (1996b) has shown that picture priming can be sensitive to size in the context of study and test tasks that required size judgment. This finding suggests that repetition priming may be sensitive to the particular processing demands of the tasks in which the stimuli appear, and may not be a fixed indicator of the stimulus attributes germane to perceptual processing in general. However, this issue has thus far received little attention in the repetition-priming literature.

Purity of Repetition-Priming Measures

Despite the large body of evidence suggesting a distinction between repetition priming and forms of explicit memory, Perruchet and Baveux (1989) demonstrated that performance on certain repetition-priming measures, such as word fragment completion and perceptual clarification (participants identified words that were embedded within a gradually disappearing mask), was correlated with performance on explicit memory tasks, whereas performance on other repetitionpriming measures, such as perceptual identification and anagram solution, was not. On this basis, Perruchet and Baveux made a distinction between two classes of repetition-priming tasks. Some tasks may be successfully solved through the use of systematic, controlled procedures (strategic tasks, which may correlate with explicit memory performance). For other tasks, however, the solution seems to pop out from a diffuse, undirected exploration (nonstrategic tasks).

The observations made by Perruchet and Baveux (1989) highlight an important issue in repetition-priming research: To what extent does a particular measure of repetition priming reflect what it is intended to measure? Among measures of perceptual repetition priming, for instance, most include at least some conceptual component. It can be difficult to completely separate facilitation based on perceptual features from facilitation based on meaning when the stimuli themselves have both perceptual and semantic qualities. To control for these effects, some researchers have chosen to use novel stimuli to eliminate the possibility of semantic information contributing to perceptual repetition-priming performance (e.g., Musen & Treisman, 1990; Schacter et al., 1990a).

Likewise, some measures of repetition priming may be open to the use of the same strategies used to make explicit memory decisions. In some instances, the only distinction between a measure of perceptual repetition priming and a measure of explicit memory may be a change in the test instruction. For instance, stem completion and cued recall differ only in that participants in a stem completion task are instructed to complete the stems with the first word that comes to mind, rather than with a previously presented word. Of course, this does not imply that participants are always using explicit memory strategies to perform priming tasks, but it does highlight some of the difficulties in establishing a pure measure of repetition priming in a normal population.

Summary

As should be evident from this survey of repetition priming research, one of the hallmark characteristics of repetition priming is that it is robust. Repetition-priming effects have been demonstrated in patients with neurological disorders as well as in normal populations, and at a wide range of intervals across the lifespan, using a vast array of different measures. Although repetition priming research is still a relatively new science, researchers have amassed a rich data store of knowledge on the subject. Despite the fact that much of what we know about repetition priming comes from research designed to address the multiple-systems versus processing debate, a focus on supporting one view over the other may not be the most fruitful path toward achieving an understanding of the mechanisms that underlie repetition priming and other memory phenomena. The more integrated perspective afforded by both views may allow for a more comprehensive understanding of these mechanisms.

Conclusions

In this research paper, we tried to provide an historical overview of the major theoretical and empirical advances in our understanding of semantic memory, semantic priming, and repetition priming. Significant progress has been made on both the theoretical and the empirical fronts in each domain.

Although the concept of semantic memory remains heuristic, semantic memory is no longer a coherent domain of inquiry, as nearly all of the phenomena originally associated with semantic memory, such as how word meanings are mentally represented and how language is understood, have become separate research endeavors. The most important recent theoretical advances have been the development of distributed network models and high-dimensional spatial models of knowledge representation. A promising direction for future research is to explore these models in more depth. Another important target of future research is the connection between perceptual systems and semantic representations. This issue is fundamentally important, yet it has received relatively little attention from experimental psychologists (but see A. S. Meyer, Sleiderink, & Levelt, 1998; Tannenhaus, Spivey-Knowlton, Eberhard, & Sedivy, 1995; Zelinsky & Murphy, 2000).

Semantic priming continues to be actively investigated more than 30 years after its discovery by D. E. Meyer and Schvaneveldt (1971). In our opinion, the most important new finding in the past decade is semantic priming over very long lags (e.g., S. Becker et al., 1997; Joordens & Becker, 1997). There are reasons to question the reliability of long-term semantic priming, but if these doubts can be put to rest with independent replications, long-term semantic priming will effectively rule out all existing models of semantic priming with the exception of distributed network learning models. Several other topics are in need of additional empirical or theoretical work, including (a) the variables that control semantic matching, (b) semantic priming for different types of semantic relations, (c) backward priming, and (d) augmenting models of priming with attentional processes that would allow the models to account for prime-task and global context effects.

The evolution of the concept of implicit memory has been one of the most important developments in the cognitive sciences in the past 20 years. Research on implicit memory in general, and repetition priming in particular, continues unabated. We believe that the most important goal for future research is to understand the mechanisms underlying repetition priming. Past research on implicit memory has been dominated by empirical issues or broad theoretical themes rather than by attempts to understand the mental representations and processes involved in repetition priming. A huge literature has now been amassed on various kinds of priming effects; now researchers need to attempt to divine the mechanisms responsible for these effects. An essential component of this endeavor will be the development of models of the sensory, perceptual, and cognitive processes responsible for repetition priming. In our opinion, the recent development of such models (e.g., Ratcliff & McKoon, 1997) represents a major step forward and is an extremely important direction for future research on implicit memory.

Bibliography:

1. Abrams, R. L., & Greenwald, A. G. (2000). Parts outweigh the whole (word) in unconscious analysis of meaning. Psychological Science, 11, 118–124.
2. Anderson, R. (1976). Language, memory, and thought .Cambridge, MA: Harvard University Press.
3. Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard University Press.
4. Balota, D. A., & Lorch, R. F. (1986). Depth of automatic spreading activation: Mediated priming effects in pronunciation but not in lexical decision. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 336–345.
5. Becker, C. A. (1980). Semantic context effects in visual word recognition: An analysis of semantic strategies. Memory & Cognition, 8, 493–512.
6. Becker, S., Moscovitch, M., Behrmann, M., & Joordens, S. (1997). Long–term semantic priming: A computational account and empirical evidence. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1059–1082.
7. Besner, , & Stolz, J. A. (1999). What kind of attention modulates the Stroop effect? Psychonomic Bulletin & Review, 6, 99–104.
8. Besner, D., Twilley, L., McCann, R. S., & Seergobin, K. (1990). On the association between connectionism and data: Are a few words necessary? Psychological Review, 97, 432–446.
9. Biederman, I., & Cooper, E. E. (1991a). Priming contour-deleted images: Evidence for intermediate representations in visual object recognition. Cognitive Psychology, 23, 393–419.
10. Biederman, I., & Cooper, E. E. (1991b). Evidence for complete translational and reflectional invariance in visual object priming. Perception, 20, 585–593.
11. Biederman, I., & Cooper, E. E. (1992). Size invariance in visual object priming. Journal of Experimental Psychology: Human Perception and Performance, 18, 121–133.
12. Blaxton, T. A. (1989). Investigating dissociations among memory measures: Support for a transfer-appropriate processing framework. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 657–668.
13. Bowers, J. S. (1999). Priming is not all bias: Commentary on Ratcliff and McKoon (1997). Psychological Review, 106, 582– 596.
14. Burgess, C., & Lund, K. (2000). The dynamics of meaning in memory. In E. Dietrich & B. Arthur (Eds.), Cognitive dynamics: Conceptual and representational change in humans and machines (pp. 117–156). Mahwah, NJ: Erlbaum.
15. Carroll, M. V., Byrne, B., & Kirsner, K. (1985). Autobiographical memory and perceptual learning: A developmental study using picture recognition, naming latency, and perceptual identification. Memory & Cognition, 13, 273–279.
16. Cave, C. B. (1997). Very long-lasting priming in picture naming. Psychological Science, 8, 322–325.
17. Cave, C. B., Bost, P. R., & Cobb, R. E. (1996). Effects of color and pattern on implicit and explicit picture memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 639–653.
18. Cave, C. B., & Squire, L. R. (1992). Intact and long-lasting repetition priming in amnesia. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 509–520.
19. Cheesman, J., & Merikle, P. (1984). Priming with and without awareness. Perception & Psychophysics, 36, 387–395.
20. Church, B. A., & Schacter, D. L. (1994). Perceptual specificity of auditory priming: Implicit memory for voice intonation and fundamental frequency. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 521–533.
21. Chwilla, D. J., Hagoort, P., & Brown, C. M. (1998). The mechanism underlying backward priming in a lexical decision task: Spreading activation versus semantic matching. Quarterly Journal of Experimental Psychology, 51A, 531–560.
22. Cohen, J. (1977). Statistical power analysis for the behavioral sciences. New York: Academic Press.
23. Cohen, N. J., & Squire, L. R. (1980). Preserved learning and retention of pattern-analyzing skill in amnesia: Dissociation of knowing how and knowing that. Science, 210, 207–210.
24. Collins, A. M., & Loftus, E. F. (1975). Aspreading-activation theory of semantic processing. Psychological Review, 82, 407–428.
25. Collins, A. M., & Quillian, M. R. (1969). Retrieval time from semantic memory. Journal of Verbal Learning and Verbal Behavior, 8, 240–247.
26. Collins, A. M., & Quillian, M. R. (1972). Experiments on semantic memory and language comprehension. In L. W. Gregg (Ed.), Cognition in learning and memory (pp. 117–147). New York: Wiley.
27. Conrad, C. (1972). Cognitive economy in semantic memory. Journal of Experimental Psychology, 92, 149–154.
28. Cooper, L. A., Schacter, D. L., Ballesteros, S., & Moore, C. (1992). Priming and recognition of transformed three-dimensional objects: Effects of size and reflection. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 43–57.
29. de Groot, A. M. (1983). The range of automatic spreading activation in word priming. Journal of Verbal Learning and Verbal Behavior, 22, 417–436.
30. de Groot, A. M. (1984). Primed lexical decision: Combined effects of the proportion of related prime-target pairs and the stimulusonset asynchrony of prime and target. Quarterly Journal of Experimental Psychology, 36A, 253–280.
31. den Heyer, K., Briand, K., & Dannenbring, G. L. (1983). Strategic factors in a lexical-decision task: Evidence for automatic and attention-driven processes. Memory & Cognition, 11, 374– 381.
32. den Heyer, K., Briand, K., & Smith, L. (1985).Automatic and strategic effects in semantic priming:An examination of Becker’s verification model. Memory & Cognition, 13, 228–232.
33. Dosher, B. A., & Rosedale, G. (1989). Integrated retrieval cues as a mechanism for priming in retrieval from memory. Journal of Experimental Psychology: General, 118, 191–211.
34. Draine, S. C., & Greenwald, A. G. (1998). Replicable unconscious semantic priming. Journal of Experimental Psychology: General, 127, 286–303.
35. Drummey, A. B., & Newcombe, N. (1995). Remembering versus knowing the past: Children’s explicit and implicit memories for pictures. Journal of Experimental Child Psychology, 59, 549–565.
36. Durgunoglu, A. Y. (1988). Repetition, semantic priming, and stimulus quality: Implications for the interactive-compensatory reading model. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 590–603.
37. Easton, R. D., Srinivas, K., & Greene, A. J. (1997). Do vision and haptics share common representations? Implicit and explicit memory within and between modalities. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 153–163.
38. Farah, M. J., & McClelland, J. L. (1991). A computational model of semantic memory impairment: Modality specificity and emergent category specificity. Journal of Experimental Psychology: General, 120, 339–357.
39. Fischler, I. (1977a). Associative facilitation without expectancy in a lexical decision task. Journal of Experimental Psychology: Human Perception and Performance, 3, 18–26.
40. Fischler, I. (1977b). Semantic facilitation without association in a lexical decision task. Memory & Cognition, 5, 335–339.
41. Fodor, J. A., & Pylyshyn, Z. W. (1988). Connectionism and cognitive architecture: A critical analysis. Cognition, 28, 3–71.
42. Forster, K. I. (1981). Priming and the effects of sentence and lexical contexts on naming time: Evidence for autonomous lexical processing. Quarterly Journal of Experimental Psychology, 33A, 465–495.
43. Franks, J. J., Bilbrey, C. W., Lien, K. G., & McNamara, T. P. (2000). Transfer-appropriate processing (TAP) and repetition priming. Memory & Cognition, 28, 1140–1151.
44. Friedrich, F. J., Henik, A., & Tzelgov, J. (1991). Automatic processes in lexical access and spreading activation. Journal of Experimental Psychology: Human Perception and Performance, 17, 792–806.
45. Gabrieli, J. D. E., Fleischman, D. A., Keane, M. M., Reminger, S. L., & Morrell, F. (1995). Double dissociation between memory systems underlying explicit and implicit memory in the human brain. Psychological Science, 6, 76–82.
46. Gabrieli, D. E., Vaidya, C. J., Stone, M., Francis, W. S., Thompson Schill, S. L., Fleischman, D. A., Tinklenberg, J. R., Yesavage, J. A., & Wilson, R. S. (1999). Convergent behavioral and neuropsychological evidence for a distinction between identification and production forms of repetition priming. Journal of Experimental Psychology: General, 128, 479–498.
47. Gillund, G., & Shiffrin, R. M. (1984). A retrieval model for both recognition and recall. Psychological Review, 91, 1–67.
48. Glass, A. L., & Holyoak, K. J. (1974). Alternative conceptions of semantic theory. Cognition, 3, 313–339.
49. Graf, P., & Mandler, G. (1984). Activation makes words more accessible, but not necessarily more retrievable. Journal of Verbal Learning and Verbal Behavior, 23, 553–568.
50. Graf, P., & Ryan, L. (1990). Transfer-appropriate processing for implicit and explicit memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 978–992.
51. Graf, P., Shimamura, A. P., & Squire, L. R. (1985). Priming across modalities and priming across category levels: Extending the domain of preserved function in amnesia. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11, 386–396.
52. Graf, P., Squire, L. R., & Mandler, G. (1984). The information that amnesic patients do not forget. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 164–178.
53. Greenbaum, J. L., & Graf, P. (1989). Preschool period development of implicit and explicit remembering. Bulletin of the Psychonomic Society, 27, 417–420.
54. Greenwald, A. G., Draine, S. C., & Abrams, R. L. (1996). Three cognitive markers of unconscious semantic activation. Science, 273, 1699–1702.
55. Greenwald, A. G., Klinger, M. R., & Schuh, E. S. (1995). Activation by marginally perceptible (“subliminal”) stimuli: Dissociation of unconscious from conscious perception. Journal of Experimental Psychology: General, 124, 22–42.
56. Hayes, B. K., & Hennessy, R. (1996). The nature and development of nonverbal implicit memory. Journal of Experimental Child Psychology, 63, 22–43.
57. Hebb, D. O. (1949). The organization of behavior. New York: Wiley.
58. Henik, A., Friedrich, F. J., & Kellogg, W. A. (1983). The dependence of semantic relatedness effects upon prime processing. Memory & Cognition, 11, 366–373.
59. Henik, A., Friedrich, F. J., Tzelgov, J., & Tramer, S. (1994). Capacity demands of automatic processes in semantic priming. Memory & Cognition, 22, 157–168.
60. Hess, D. J., Foss, D. J., & Carroll, P. (1995). Effects of global and local context on lexical processing during language comprehension. Journal of Experimental Psychology: General, 124, 62–82.
61. Hinton, G. E., & Shallice, T. (1991). Lesioning an attractor network: Investigations of acquired dyslexia. Psychological Review, 98, 74–95.
62. Hintzman, D. L. (1986). “Schema abstraction” in a multiple-trace memory model. Psychological Review, 93, 411–428.
63. Holender, D. (1986). Semantic activation without conscious identification in dichotic listening, parafoveal vision, and visual masking: A survey and appraisal. Behavioral and Brain Sciences, 9, 1–66.
64. Holyoak, K. J., & Glass, A. L. (1975). The role of contradictions and counterexamples in the rejection of false sentences. Journal of Verbal Learning and Verbal Behavior, 14, 215–239.
65. Humphreys, G. W., & Quinlan, P. T. (1987). Normal and pathological processes in visual object constancy. In G. W. Humphreys & M. J. Riddoch (Eds.), Visual object processing: A cognitive neuropsychological approach (pp. 43–105). Hillsdale, NJ: Erlbaum.
66. Jackson, A., & Morton, J. (1984). Facilitation of auditory word recognition. Memory & Cognition, 12, 568–574.
67. Jacoby, L. L. (1991). A process dissociation framework: Separating automatic from intentional uses of memory. Journal of Memory and Language, 30, 513–541.
68. Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General, 110, 306–340.
69. Joordens, S., & Becker, S. (1997). The long and short of semantic priming effects in lexical decision. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1083–1105.
70. Joordens, S., & Besner, D. (1992). Priming effects that span an intervening unrelated word: Implications for models of memory representation and retrieval. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 483–491.
71. Kahan, T. A., Neely, J. H., & Forsythe, W. J. (1999). Dissociated backward priming effects in lexical decision and pronunciation tasks. Psychonomic Bulletin & Review, 6, 105–110.
72. Kawamoto, A. H., Farrar, W. T., & Kello, C. T. (1994). When two meanings are better than one: Modeling the ambiguity advantage using a recurrent distributed network. Journal of Experimental Psychology: Human Perception and Performance, 20, 1233–1247.
73. Keane, M. M., Gabrieli, J. D. E., Mapstone, H. C., Johnson, K. A., & Corkin, S. (1995). Double dissociation of memory capacities after bilateral occipital-lobe or medial temporal lobe lesions. Brain, 118, 1129–1148.
74. Keefe, D. E., & Neely, J. H. (1990). Semantic priming in the pronunciation task: The role of prospective prime-generated expectancies. Memory & Cognition, 18, 289–298.
75. Kirsner, K., & Speelman, C. (1996). Skill acquisition and repetition priming: One principle, many processes? Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 563–575.
76. Klinger, M. R., Burton, P. C., & Pitts, G. S. (2000). Mechanisms of unconscious priming: I. Response competition, not spreading activation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, 441–455.
77. Koriat, A. (1981). Semantic facilitation in lexical decision as a function of prime-target association. Memory & Cognition, 9, 587–598.
78. Landauer, T. K. (1998). Learning and representing verbal meaning: The Latent Semantic Analysis Theory. Current Directions in Psychological Science, 7, 161–164.
79. Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato’s problem: The latent semantic analysis theory of acquisition, induction, and representation of knowledge. Psychological Review, 104, 211–240.
80. Livesay, K., & Burgess, C. (1998). Mediated priming in highdimensional semantic space: No effect of direct semantic relationships or co-occurrence. Brain and Cognition, 37, 102–105.
81. Logan, G. D. (1988). Toward an instance theory of automatization. Psychological Review, 95, 492–527.
82. Logan, G. D. (1990). Repetition priming and automaticity: Common underlying mechanisms? Cognitive Psychology, 22, 1–35.
83. Lucas, M. (2000). Semantic priming without association: A metaanalytic review. Psychonomic Bulletin & Review, 7, 618–630.
84. Lund, K., Burgess, C., & Audet, C. (1996). Dissociating semantic and associative word relationships using high-dimensional semantic space. Proceedings of the Eighteenth Annual Conference of the Cognitive Science Society, 18, 603–608.
85. Lupker, S. J. (1984). Semantic priming without association: A second look. Journal of Verbal Learning and Verbal Behavior, 23, 709–733.
86. Marcel, A. (1983). Conscious and unconscious perception: Experiments on visual masking and word recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 108–115.
87. Marsolek, C. J., Kosslyn, S. M., & Squire, L. R. (1992). Formspecific visual priming in the right cerebral hemisphere. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 492–508.
88. Masson, M. E. J. (1991). Adistributed memory model of context effects in word identification. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 233–263). Hillsdale, NJ: Erlbaum.
89. Masson, M. E. J. (1995). A distributed memory model of semantic priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 3–23.
90. McClelland, J. L., & Rumelhart, D. E. (1985). Distributed memory and the representation of general and specific information. Journal of Experimental Psychology: General, 114, 159–188.
91. McClelland, J. L., & Rumelhart, D. E. (Eds.). (1986). Parallel distributed processing: Explorations in the microstructure of cognition: Vol. 2. Psychological and biological models. Cambridge, MA: MIT Press.
92. McCloskey, M., & Glucksberg, S. (1979). Decision processes in verifying category membership statements: Implications for models of semantic memory. Cognitive Psychology, 11, 1–37.
93. McDermott, K. B., & Roediger, H. L. (1994). Effects of imagery on perceptual implicit memory tests. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1379– 1390.
94. McKoon, G., & Ratcliff, R. (1992). Spreading activation versus compound cue accounts of priming: Mediated priming revisited. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1155–1172.
95. McKoon, G., & Ratcliff, R. (1995). Conceptual combinations and relational contexts in free association and in priming in lexical decision and naming. Psychonomic Bulletin & Review, 2, 527–533.
96. McKoon, G., Ratcliff, R., & Dell, G. S. (1986). A critical evaluation of the semantic-episodic distinction. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 295– 306.
97. McNamara, T. P. (1992a). Priming and constraints it places on theoriesofmemoryandretrieval.PsychologicalReview,99,650–662.
98. McNamara, T. P. (1992b). Theories of priming: I. Associative distance and lag. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1173–1190.
99. McNamara, T. P., & Altarriba, J. (1988). Depth of spreading activation revisited: Semantic mediated priming occurs in lexical decisions. Journal of Memory and Language, 27, 545–559.
100. McRae, K., & Boisvert, S. (1998). Automatic semantic similarity priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 558–572.
101. McRae, K., de Sa, V. R., & Seidenberg, M. S. (1997). On the nature and scope of featural representations of word meaning. Journal of Experimental Psychology: General, 126, 99–130.
102. Meyer, A. S., Sleiderink, A. M., & Levelt, W. J. M. (1998). Viewing and naming objects: Eye movements during noun phrase production. Cognition, 66, B25–B33.
103. Meyer, D. E. (1970). On the representation and retrieval of stored semantic information. Cognitive Psychology, 1, 242–300.
104. Meyer, D. E., & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words: Evidence of a dependence between retrieval operations. Journal of Experimental Psychology, 90, 227–234.
105. Mitchell, D. B. (1989). How many memory systems? Evidence from aging. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 31–49.
106. Mitchell, D. B. (1993). Implicit and explicit memory for pictures: Multiple views across the lifespan. In P. Graf & M. E. J. Masson (Eds.),Implicitmemory:Newdirectionsincognition,development, and neuropsychology (pp. 171–190). Hillsdale, NJ: Erlbaum.
107. Mitchell, D. B., & Brown, A. S. (1988). Persistent repetition priming in picture naming and its dissociation from recognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 213–222.
108. Monsell, S. (1985). Repetition and the lexicon. In A. W. Ellis (Ed.), Progress in the psychology of language (pp. 147–196). London: Erlbaum.
109. Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16, 519–533.
110. Morton, J. (1969). Interaction of information in word recognition. Psychological Review, 76, 165–178.
111. Moss, H. E., Hare, M. L., Day, P., & Tyler, L. K. (1994). A distributed memory model of the associative boost in semantic priming. Connection Science: Journal of Neural Computing, Artificial Intelligence, and Cognitive Research, 6, 413–427.
112. Moss, H. E., Ostrin, R. K., Tyler, L. K., & Marslen Wilson, W. D. (1995). Accessing different types of lexical semantic information: Evidence from priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 863–883.
113. Murdock, B. B. (1982). A theory for the storage and retrieval of item and associative information. Psychological Review, 89, 609–626.
114. Musen, G., & Treisman, A. (1990). Implicit and explicit memory for visual patterns. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 127–137.
115. Neely, J. H. (1977). Semantic priming and retrieval from lexical memory: Roles of inhibitionless spreading activation and limited-capacity attention. Journal of Experimental Psychology: General, 106, 226–254.
116. Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading: Visual word recognition (pp. 264–336). Hillsdale, NJ: Erlbaum.
117. Neely, J. H., & Kahan, T. A. (2001). Is semantic activation automatic? A critical re-evaluation. In H. L. Roediger III, J. S. Nairne, I. Neath, & A. M. Surprenant (Eds.), The nature of remembering: Essays in honor of Robert G. Crowder (pp. 69–93). Washington, DC: American Psychological Association.
118. Neely, J. H., & Keefe, D. E. (1989). Semantic context effects on visual word processing: A hybrid prospective-retrospective processing theory. In G. H. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (Vol. 24, pp. 207–248). New York: Academic Press.
119. Neely, J. H., Keefe, D. E., & Ross, K. L. (1989). Semantic priming in the lexical decision task: Roles of prospective primegenerated expectancies and retrospective semantic matching. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 1003–1019.
120. Nelson, D. L., McEvoy, C. L., & Schreiber, T. A. (1991). The University of South Florida word association, rhyme, and word fragment norms. Unpublished manuscript.
121. Oscar-Berman, M. (1984). Comparative neuropsychology and alcoholic Korsakoff disease. In L. R. Squire & N. Butters (Eds.), Neuropsychology of Memory (pp. 194–202). New York: Guilford Press.
122. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana: University of Illinois Press.
123. Perruchet, P., & Baveux, P. (1989). Correlational analyses of explicit and implicit memory performance. Memory & Cognition, 17, 77–86.
124. Peterson, R. R., & Simpson, G. B. (1989). Effect of backward priming on word recognition in single-word and sentence contexts. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 1020–1032.
125. Plaut, D. C., & Booth, J. R. (2000). Individual and developmental differences in semantic priming: Empirical and computational support for a single-mechanism account of lexical processing. Psychological Review, 107, 786–823.
126. Plaut, D. C., McClelland, J. L., Seidenberg, M. S., & Patterson, K. (1996). Understanding normal and impaired word reading: Computational principles in quasi-regular domains. Psychological Review, 103, 56–115.
127. Posner, M. I., & Snyder, C. R. R. (1975). Attention and cognitive control. In R. L. Solso (Ed.), Information processing and cognition (pp. 55–85). Hillsdale, NJ: Erlbaum.
128. Quillian, M. R. (1967). Word concepts: A theory and simulation of some basic semantic capabilities. Behavioral Science, 12, 410–430.
129. Quillian, M. R. (1969). The teachable language comprehender. Communications of the Association for Computing Machinery, 12, 459–476.
130. Rappold, V. A., & Hashtroudi, S. (1991). Does organization improve priming? Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 103–114.
131. Ratcliff, R., & McKoon, G. (1981). Does activation really spread? Psychological Review, 88, 454–462.
132. Ratcliff, R., & McKoon, G. (1982). Speed and accuracy in the processing of false statements about semantic information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 8, 16–36.
133. Ratcliff, R., & McKoon, G. (1988). A retrieval theory of priming in memory. Psychological Review, 95, 385–408.
134. Ratcliff, R., & McKoon, G. (1996). Bias effects in implicit memory tasks. Journal of Experimental Psychology: General, 125, 403–421.
135. Ratcliff, R., & McKoon, G. (1997). A counter model for implicit priming in perceptual word identification. Psychological Review, 104, 319–343.
136. Ratcliff, R., & McKoon, G. (2000). Modeling the effects of repetition and word frequency in perceptual identification. Psychonomic Bulletin & Review, 7, 713–717.
137. Richardson-Klavehn, A., & Bjork, R. A. (1988). Measures of memory. Annual Review of Psychology, 39, 475–543.
138. Riddoch, M. J., & Humphreys, G. W. (1987). Picture naming. In G. W. Humphreys & M. Riddoch (Eds.), Visual object processing: A cognitive neuropsychological approach (pp. 107–143). Hillsdale, NJ: Erlbaum.
139. Rips, L. J., Shoben, E. J., & Smith, E. E. (1973). Semantic distance and the verification of semantic relations. Journal of Verbal Learning and Verbal Behavior, 12, 1–20.
140. Roediger, H. L., III. (1990). Implicit memory: Retention without remembering. American Psychologist, 45, 1043–1056.
141. Roediger, H. L., III, & Blaxton, T. A. (1987). Effects of varying modality, surface features, and retention interval on priming in word-fragment completion. Memory & Cognition, 15, 379– 388.
142. Roediger, H. L., III, Weldon, M. S., & Challis, B. H. (1989). Explaining dissociations between implicit and explicit measures of retention: A processing account. In H. L. Roediger III & F. I. M. Craik (Eds.), Varieties of memory and consciousness: Essays in honour of Endel Tulving (pp. 2–41). Hillsdale, NJ: Erlbaum.
143. Rosenblatt, F. (1962). Principles of neurodynamics. New York: Spartan.
144. Rueckl, J. G., Mikolinski, M., Raveh, M., Miner, C. S., & Mars, F. (1997). Morphological priming, fragment completion, and connectionist networks. Journal of Memory and Language, 36, 382–405.
145. Rumelhart, D. E., & McClelland, J. L. (Eds.). (1986). Parallel distributed processing: Explorations in the microstructure of cognition: Vol. 1. Foundations. Cambridge, MA: MIT Press.
146. Schacter, D. L. (1990). Perceptual representation systems and implicit memory: Toward a resolution of the multiple memory systems debate. Annals of the New York Academy of Sciences, 608, 543–571.
147. Schacter, D. L. (1992). Priming and multiple memory systems: Perceptual mechanisms of implicit memory. Journal of Cognitive Neuroscience, 4, 244–256.
148. Schacter, D. L., Church, B., & Bolton, E. (1995). Implicit memory in amnesic patients: Impairment of voice-specific priming. Psychological Science, 6, 20–25.
149. Schacter, D. L., Church, B., & Treadwell, J. (1994). Implicit memory in amnesic patients: Evidence for spared auditory priming. Psychological Science, 5, 20–25.
150. Schacter, D. L., Cooper, L. A., & Delaney, S. M. (1990a). Implicit memory for unfamiliar objects depends on access to structural descriptions. Journal of Experimental Psychology: General, 119, 5–24.
151. Schacter, D. L., Cooper, L. A., & Delaney, S. M. (1990b). Implicit memory for visual objects and the structural description system. Bulletin of the Psychonomic Society, 28, 367–372.
152. Seidenberg, M. S., & McClelland, J. L. (1989). Adistributed, developmental model of word recognition and naming. Psychological Review, 96, 523–568.
153. Seidenberg, M. S., Waters, G. S., Sanders, M., & Langer, P. (1984). Pre- and postlexical loci of contextual effects on word recognition. Memory & Cognition, 12, 315–328.
154. Sharkey, A. J., & Sharkey, N. E. (1992). Weak contextual constraints in text and word priming. Journal of Memory and Language, 31, 543–572.
155. Shelton, J. R., & Martin, R. C. (1992). How semantic is automatic semantic priming? Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1191–1210.
156. Shimamura, A. P. (1986). Priming effects in amnesia: Evidence for a dissociable memory function. Quarterly Journal of Experimental Psychology, 38A, 619–644.
157. Shimamura, A. P. (1989). Disorders of memory: The cognitive science perspective. In F. Boller & J. Grafman (Eds.), Handbook of neuropsychology (pp. 35–73). Amsterdam: Elsevier Science.
158. Shimamura, A. P. (1990). Forms of memory: Issues and directions. In J. L. McGaugh, N. M. Weinberger, & G. Lynch (Eds.), Brain organization and memory: Cells, systems, and circuits (pp. 159– 173). New York: Oxford University Press.
159. Shimamura, A. P. (1993). Neuropsychological analyses of implicit memory: History, methodology and theoretical interpretations. In P. Graf & M. E. Masson (Eds.), Implicit memory: New directions in cognition, development, and neuropsychology (pp. 265– 285). Hillsdale, NJ: Erlbaum.
160. Sloman, S. A., Hayman, C. A. G., Ohta, N., Law, J., & Tulving, E. (1988). Forgetting in primed fragment completion. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 223–239.
161. Smith, E. E. (1978). Theories of semantic memory. In W. K. Estes (Ed.), Handbook of learning and cognitive processes: Vol. 4. Linguistic functions in cognitive theory (pp. 1–56). Potomic, MD: Erlbaum.
162. Smith, E. E., Shoben, E. J., & Rips, L. J. (1974). Structure and process in semantic memory: A featural model for semantic decisions. Psychological Review, 81, 214–241.
163. Smith, M. C. (1979). Contextual facilitation in a letter search task depends on how the prime is processed. Journal of Experimental Psychology: Human Perception and Performance, 5, 239–251.
164. Smith, M. C., Theodor, L., & Franklin, P. E. (1983). The relationship between contextual facilitation and depth of processing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 697–712.
165. Squire, L. R. (1986). Mechanisms of memory. Science, 232, 1612–1619.
166. Squire, L. R. (1987). Memory and brain. New York: Oxford University Press.
167. Squire, L. R. (1992). Memory and the hippocampus: A synthesis from findings with rats, monkeys, and humans. Psychological Review, 99, 195–231.
168. Squire, L. R., Zola-Morgan, S., Cave, C. B., Haist, F., Musen, G., & Suzuki, W. A. (1993). Memory: Organization of brain systems and cognition. In D. E. Meyer & S. Kornblum (Eds.), Attention and performance XIV (pp. 393–424). Cambridge, MA: MIT
169. Srinivas, K. (1996a). Contrast and illumination effects on explicit and implicit measures of memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 1123–1135.
170. Srinivas, K. (1996b). Size and reflection effects in priming: A test of transfer-appropriate processing. Memory & Cognition, 24, 441–452.
171. Stark, C. E. L., & McClelland, J. L. (2000). Repetition priming of words, pseudowords, and nonwords. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, 945–972.
172. Stolz, J. A., & Besner, D. (1996). Role of set in visual word recognition: Activation and activation blocking as nonautomatic processes. Journal of Experimental Psychology: Human Perception and Performance, 22, 1166–1177.
173. Stolz, J. A., & Besner, D. (1999). On the myth of automatic semantic activation in reading. Current Directions in Psychological Science, 8, 61–65.
174. Tannenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K., & Sedivy, J. C. (1995). Integration of visual and linguistic information in spoken language comprehension. Science, 268, 1632–1634.
175. Thompson-Schill, S. L., Kurtz, K. J., & Gabrieli, J. D. E. (1998). Effects of semantic and associative relatedness on automatic priming. Journal of Memory and Language, 38, 440–458.
176. Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381–403). New York: Academic Press.
177. Tulving, E. (1983). Elements of episodic memory. New York: Oxford University Press.
178. Tulving, E., & Schacter, D. L. (1990). Priming and human memory systems. Science, 247, 301–306.
179. Tulving, E., Schacter, D. L., & Stark, H. A. (1982). Priming effects in word-fragment completion are independent of recognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 8, 336–342.
180. Tweedy, J. R., Lapinski, R. H., & Schvaneveldt, R. W. (1977). Semantic-context effects on word recognition: Influence of varying the proportion of items presented in an appropriate context. Memory & Cognition, 5, 84–89.
181. Tyler, L. K., & Moss, H. E. (1997). Functional properties of concepts: Studies of normal and brain-damaged patients. Cognitive Neuropsychology, 14, 511–545.
182. Vaidya, C., Gabrieli, J. D. E., Keane, M. M., Monti, L. A., Gutierrez Rivas, H., & Zarella, M. M. (1997). Evidence for multiple mechanisms of conceptual priming on implicit memory tests. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1324–1343.
183. Wagenmakers, E. J. M., Zeelenberg, R., & Raaijmakers, J. G. W. (2000). Testing the counter model for perceptual identification: Effects of repetition priming and word frequency. Psychonomic Bulletin & Review, 7, 662–667.
184. Warrington, E. K., & Weiskrantz, L. (1968). New method of testing long-term retention with special reference to amnesic patients. Nature, 217, 972–974.
185. Weldon, M. S. (1991). Mechanisms underlying priming on perceptual tests. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 526–541.
186. Zelinsky, G. J., & Murphy, G. L. (2000). Synchronizing visual and language processing: An effect of object name length on eye movements. Psychological Science, 11, 125–131.