Linda A. W. Brakel 52/4

THE PSYCHOANALYTICASSUMPTION OFTHE PRIMARY PROCESS:EXTRAPSYCHOANALYTICEVIDENCE AND FINDINGS

No evidence for an assumption of a theory can be gained by dataderived from methods dependent on that theory. Three experiments,using methods independent of psychoanalysis, test the psychoanalyticposit that primary process exists as a formal mental mode distinct fromsecondary process. The three experiments, using a nonpsychoanalyticindex for primary process, test for a preponderance of primary processorganization in three areas in which Freud observed primary process:(1) in unconscious mental states and during implicit tasks; (2) in the mentalproductions of preschool children; and (3) in anxiety states, as these aretypically associated with unconscious conflict. All three experimentsshow significant results in favor of the primary process. Further, the threeexperiments taken together, because they account for seemingly dis-parate data, lend further credence to the original assumption. These pos-itive results suggest that primary process may be more important thaneven Freud suspected. Perhaps it is the basic mental organization in manynonhuman mammals and some birds. Primary process organization mayalso underlie aspects of such basic psychological operations as general-izations in conditioning and assessments-in-action, as opposed to judg-ments proper, about how one would act. Finally, primary process mayplay a key role in drive activity. Three types of experiments are proposedto test these far-reaching applications of the primary process concept.

E very major scientific theory has at its core a number of basicassumptions. Initially, for a science to progress, these basic assump-

tions are taken as true, without test, and serve as the foundation for theobservations explained and the research conducted. It is only at a laterphase in the science’s development that the assumptions of its majortheories must be tested. For such tests to be valid they must use methods

japa

that do not presuppose the truth of the assumptions. When basic assump-tions can gain this sort of independent support, confidence in the validityof the science is greatly enhanced. Work in our laboratory,1 by provid-ing independent evidence for two central psychoanalytic assumptions,has given reason for optimism regarding the status of psychoanalytictheory. The two basic psychoanalytic assumptions we have investi-gated are (1) that there exist unconscious, as well as conscious, mentalprocesses and contents, and (2) that there exist two distinct formal orga-nizational structures of these conscious and unconscious mentalprocesses and contents—the primary process form or mode of organizingand the secondary process mode of organizing. When these assumptionsare taken as true, cogent explanations can be made for a very large rangeof observed behavioral phenomena—including those, like psychologicalsymptoms, that seem incoherent under other explanations. Here is anexample that can be readily explained by taking both of the abovepsychoanalytic assumptions as true:

Mrs. M was a middle-aged woman with the history of a chronic inter-mittent psychosis. At the time I f irst saw her, she was hallucinating,screaming with pain, and shouting, “The delivery of the head andshoulders is killing me,” as she forcefully pushed a hard plastic bottleof Head and Shoulders shampoo against her pelvis. Mrs. M, whohad given birth to three children via normal deliveries decades before,now faced an impending surgery for an advanced and painful gyneco-logical malignancy.

The explanation for Mrs. M’s behavior, with the assumptions taken astrue, goes as follows. Mrs. M’s anxiety about her life-threatening dis-

L i n d a A . W. B r a k e l

1132

Associate Clinical Professor Department of Psychiatry, University of Michigan;faculty, Michigan Psychoanalytic Institute.

A version of the first sections of the paper was presented at the Austrian Academyof Sciences Conference on Psychoanalysis as an Interdisciplinary Science, Vienna,December 2001, and has been published in German in Psychoanalyse im DialogenWissenschaften, ed P. Giampieri-Deutsch (Stuttgart: Kohlhammer, 2004). Versions ofthis paper were also presented to the American Psychological Association (Division39), Minneapolis, April 2003; the American Psychoanalytic Association, Boston, June2003; as the Fisher Lecture, New York, May 2003; and to the Mind, Brain, Culture,and Consciousness Group at Yale University, April 2004. A much briefer version waspresented to the Neuro-Psychoanalysis Congress, New York, July 2003, and will beincluded in those proceedings. Submitted for publication August 14, 2003.

1The research referred to in this paper was conducted in the Ormond and HazelHunt and Family Memorial Laboratory, directed by Howard Shevrin. The laboratorywas founded and supported by gifts from Robert H. Berry, with continuing supportfrom the Smart Family Foundation.

ease and upcoming surgery occasioned a recurrence of her psychosis.This was marked by regression to primary process thinking in whichshe wished that her pain was caused by the normal delivery of the headand shoulders of babies, rather than the pain of her very abnormaltumor, which would soon cause more pain in the impending surgery.The primary process thinking that underlay this—equating the plasticHead and Shoulders shampoo bottle with the head and shoulders ofbabies—was unconscious, as was the wish that she be delivering agrowing baby again, instead of suffering from a growing cancer.

It is striking how easily this otherwise bizarre behavior can be wellunderstood with the assumptions of unconscious content and processes,and of a primary process formal mental organization, taken as true. Butalthough we can make sense of Mrs. M’s material with the assump-tions in place, since we have taken these assumptions as true we can-not use Mrs. M’s now sensible material to count as evidence for theassumptions themselves. In fact, we have done nothing to gain evidencefor these assumptions . . . yet.

But the story does not end here. Using independent methodsthat do not presuppose the truth of these two assumptions, we havebeen able to gain evidence for them. Regarding the assumption ofunconscious mental contents and processes, I will remark only brieflythat Shevrin and his colleagues have had much success. Variousstudies have demonstrated the following: (1) unconscious perceptionand cognition (Shevrin and Fisher 1967; Shevrin and Fritzler 1968b;Shevrin, Smith, and Fritzler 1971; Snodgrass, Shevrin, and Kopka1993; Brakel et al. 2000; Bernat, Shevrin, and Snodgrass 2001);(2) unconscious memory (Bunce et al. 2001); (3) unconscious con-ditioning (Wong, Shevrin, and Williams 1994; Wong et al. 1997, inpress); (4) unconscious learning (Wong et al. 1997; Bunce et al. 1999;Bernat, Shevrin, and Snodgrass 2001); (5) unconscious af fect(Bernat, Bunce, and Shevrin 2001; Shevrin et al. 2004); and (6) un-conscious conf lict (Shevrin et al. 1992, 1996). In all of these studiesthe subliminal method, a method quite independent of any of theassumptions of psychoanalysis, has been employed such thatstimuli can be delivered very rapidly at the objective detectionthreshold—in other words, stimuli are experimentally rendered con-sciously inaccessible.

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

1133

THE PRIMARY PROCESS

Regarding the second assumption, that of two distinct formal mentalorganizations—the primary and secondary process—I will not be brief.The remainder of this article concerns this assumption and the inde-pendent method we used to test it. I begin with a summary of Freud’sdescription of the primary process (1895, 1900, 1915, 1940).

Primary and secondary process were so named for two not unrelatedreasons. First, primary process is a developmentally earlier mode ofmental organization. Freud noted that much of the behavior and thoughtof the very young could be understood in terms of primary processorganization. Second, primary process organizes unconscious mentalprocesses, and even in adults unconscious processes precede consciousones. This is clearest in dreams, daydreams, fantasies, and neurotic andpsychotic symptoms, but occurs also in most conscious rational think-ing. The claims Freud made about these “different [primary process]laws” (Freud 1940, p. 164) can be summarized: where “the so-called‘primary process’ prevails, there is no synthesis of ideas, affects areliable to displacement, opposites are not mutually exclusive and mayeven coincide and condensation occurs as a matter of course. Also thesovereign principle . . . is that of obtaining pleasure” (A. Freud 1936,p. 7). In addition, the primary process is regarded as timeless—i.e.,without past or future, and “there are in this system no negation, nodoubt, no degrees of certainty” (Freud 1915, pp. 186–187). So these arethe basic primary process characteristics, save for those concerning thenature of the different energy and drive-cathexis. Although interesting(and controversial in some quarters), these matters will not be taken uphere. Indeed the work presented here has very little to say about thecontent aspects of primary process so familiar to psychoanalysts—con-tent reflecting its drive-dominated, preemptive, often primitive, andconflictual qualities. Instead the experimental work presented requiresproviding more detail about the formal structure of primary processmentation—the cognitive operations of primary process.

Throughout his seminal The Interpretation of Dreams, Freud(1900) talks of the two primary process mechanisms of condensationand displacement acting on latent dream material in “dream-work” soas to produce the manifest dreams we experience. Condensation anddisplacement are both related to a primary process tendency Freudbegan discussing in the Project for a Scientific Psychology (1895), “a

L i n d a A . W. B r a k e l

1134

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

compulsion to associate” (p. 338). Associations famously became thefoundation of the technical method, free association, but Freud earlyon noted some of their more obscure characteristics. He called them“superficial associations” (1900, p. 597) and “external associations”(1901, p. 278), but by 1905, in Jokes and Their Relation to the Uncon-scious, it is clear that for Freud the associations that often get con-densed and displaced are those pertaining to small, inessential aspectsof a thing (pp. 88–89). In accord with this, Rapaport, an importantpsychoanalytic theorist of the 1940s and 1950s, found that pars prototo, part-for-whole, is a frequently used primary process substitution(1951a, pp. 395, 398). Further capturing the fluid yet concrete primaryprocess use of parts or attributes of an object to stand for the whole,Rapaport (1951b) added, “Where the primary process . . . holds sway. . . everything belongs with everything that shares an attribute of it . . . ”(p. 708).

SETTING UP THE EXPERIMENTS

With Freud’s outline of the primary process assumption in mind, I willin this section describe the method independent of this assumption(and in fact independent of any psychoanalytic assumptions) that wasemployed to test it. The issue of the independence of testing methodsand variables from the assumptions tested is critical, as can be seenfrom the history of empirical investigation of the primary process.Particularly relevant to the experiments discussed here are three groupsof studies in which the primary process mode was studied empirically.In chronological order these were done by Poetzl (1917, translatedin Fisher 1960a); Fisher (1954, 1957, 1960a,b; Fisher and Paul 1959);and Shevrin and colleagues (Shevrin and Luborsky 1961; Shevrinand Stross 1962; Stross and Shevrin 1962; Shevrin and Fisher 1967;Shevrin and Fritzler 1968a,b).

Poetzl presented visual stimuli in a tachistoscope such that someparts of the stimuli were outside the participants’ conscious awareness.Poetzl then had them recall what they had been presented, recount theirsubsequent dreams, and draw pictures of these dreams. He found that“the memory images of the unconsciously registered elements . . . oftenare transformed and distorted in various ways . . . [that] resemble thosewhich Freud described as . . . condensations, displacements, symbolictransformations, etc. [primary process manifestations]” (Fisher 1960a,

1135

L i n d a A . W. B r a k e l

1136

pp. 17–18). Fisher’s work (1957; Fisher and Paul 1959) first replicatedPoetzl’s f indings, and then went beyond them by showing a contrastin the effects produced by different but simultaneously presented sub-liminal and supraliminal stimuli. Fisher (1960b) found that only thesubliminal stimuli underwent primary process type transformations.

The Poetzl and Fisher studies provided much anecdotal evidencefor primary process transformations, but they lacked certain features ofexperimental design—notably, appropriate controls and objective mea-sures. Shevrin’s studies remedied these problems. With a 1 millisecondduration for subliminal presentations (to essentially rule out consciousperception), a rebus figure (along with an appropriate non-rebus con-trol) was used as a stimulus that could elicit either primary processor secondary process associations depending on the conditions underwhich the associations were gathered. Shevrin predicted correctlywhich conditions would favor primary process transformations ofcontent (e.g., when participants were awakened in an REM sleep stage)and which would produce secondary process manifestations (whenparticipants were awakened in a non-REM sleep stage). Further, hespecified in advance what would be considered a primary processmanifestation (e.g., clang association or association to the rebus.)

Each of these groups of studies, while owing much to the previouswork, showed advances in independence from psychoanalytic assump-tions and conceptualizations. Poetzl found that he could describe hisfindings using the concepts of primary and secondary processes—con-cepts he assumed. Fisher too assumed primary and secondary processorganizational principles and used these to explain the transformationshe observed. Shevrin and colleagues took a major step in that in theirstudies the primary and secondary process organizational principleswere no longer used as explanatory concepts to account for the data;instead, prespecif ied manifestations of primary process constitutedthe data to be collected. In the studies to be summarized here we takea further step toward independence. Shevrin determined in advancewhat would count as a manifestation of primary process organization,but he did so on the basis of the psychoanalytic descriptive theory heassumed. In other words, clang associations are based on lexical displace-ments (rhymes) and rebus associations predicated on condensation—bothpsychoanalytic conceptualizations. In the experiments here, by contrast,the data collected are entirely independent of any aspect of psychoanalytictheory. We employed an experimental version of an aspect of primary

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

process mental organization, and we derived it from a cognitive psy-chology descriptive theory of categorization. (For a fuller account ofthe older studies and their relation to the current line of research, seeBrakel et al. 2000, pp. 555–557.)

Recall that in Freud’s characterizations of the primary process themechanisms of displacement and condensation, and the superficialassociations that got displaced and condensed, f igured prominently.So did operations like part-for-whole. Now recall also Rapaport’scomment (1951b): “Where the primary process . . . holds sway . . .everything belongs with everything that shares an attribute of it . . .”(p. 708). In reviewing quotes like these, it occurred to us that we couldcontrast primary vs. secondary process on the basis of different princi-ples of categorization. Although not addressed much in the psycho-analytic literature, categorization by superf icial association (i.e., byattribute) actually is pivotal among the primary process principles.Take, for example, cigars. Categorized in a secondary process fashion,cigars belong with cigarettes and pipes as smokes. Cigars, cigarettes,and pipes are judged similar on the basis of a fundamental shared rela-tionship among disparate parts: different blends, packed into differentforms, made of different materials, each consists of tobacco such thatwhen lit, smoke is produced with various effects when inhaled. A pri-mary process categorization of cigars groups them with phalluses.Whereas from a secondary process view cigars are not body partsand phalluses are not smokes, a primary process organization linkscigars and phallus and categorizes them together on the basis of one ormore of their similar attributes: cigars and phalluses are similar inthat they both have a cylindrical shape, red hot tips, and are capableof providing pleasure when inserted into appropriate orif ices. Or, toreturn to the case of Mrs. M, take shampoo bottles. Categorized in asecondary process fashion, my patient’s bottle of Head and Shoulderswould belong with other hair care items like her roommate’s bottle ofPrell. The Head and Shoulders would be simply a product name andthe shape of the bottle a contingent fact. But categorized in a primaryprocess way, her Head and Shoulders bottle, particularly when pressedagainst her pelvis, was categorized with the heads and shoulders ofMrs. M’s babies as they were passing through the birth canal. Thisis a primary process categorization. (For a more complete discussion ofthe nature of primary process categories, see Brakel 2004; Brakel andVelleman 2003.)

1137

These examples also make clear the vital role of similarity in anycapacity to categorize. Whether organized by attribute (primaryprocess) or relationship (secondary process), these categorizationsall require that for the categorizer some sort of similarity—explicitor implicit, recognized or imparted—exists among the category mem-bers. It is in this context that the work of cognitive psychologistDouglas Medin and his colleagues became important for us (Medin,Goldstone, and Gentner 1990, 1993; Gentner 1988, 1989; Gentner andRatterman 1991; Smith and Medin 1981; Murphy and Medin 1985).They described “attributional similarity” as categorization based on thematching of concrete, particular, and superficial features, and opposedthis to “relational similarity,” which is categorization based on match-ing relations among differing attributes (see Figure 1).

Clearly, Medin and his group derived their descriptive theoryof these two types of similarity independent of any psychoanalyticassumptions or preconceptions. Recognizing next that attributionalsimilarity and relational similarity as described by Medin et al. couldmap quite well onto a formal aspect of primary process organization(categorization by attribute) and onto a formal aspect of secondaryprocess organization (categorization by relationship), we reasoned thatattributional vs. relational similarity assessments could independentlyindex, and if found where predicted, stand as evidence for categoriza-tions mediated by primary process as opposed to secondary process.Thus we adapted the simple design of Medin et al. (1990) to constructthe items (see Figure 2).

L i n d a A . W. B r a k e l

1138

FIGURE 1. Attributional vs. Relational Similarity

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

There are two important things to point out about these itemsgenerally (for the entire item set, see Figure 3). First, because theitems require little background knowledge, they can be used acrossage groups (we have used them with participants as young as threeand as old as ninety or more). Second, the figures are counterbalancednot only for position, but also for their roles. Thus, a figure that is theattributional match in one item is the relational match in another with adifferent master figure, and vice versa. (This obviates item bias.)

THE CATEGORIZATION EXPERIMENTS

There were a few simple hypotheses. Where psychoanalytic theorypredicts a shift toward the primary process mode, we predicted a shifttoward attributional similarity responses. Freud reported an abundanceof primary process mentation in three circumstances: (1) in conditionsunder which unconscious influences are large, as when persons are notawake and alert, and/or when indirect tasks, like the free-associativeone, alter attentional conditions (examples include dreams and dreamstates, slips of the tongue, fantasies, and free associations); (2) early indevelopment; as seen in much of the mentation observable in youngchildren; and (3) in symptomatic behavior and in the development ofsymptoms. Consequently, we hypothesized a shift toward attribu-tional responses in these three cases: (1) when items were presented

1139

FIGURE 2. One of the Items

unconsciously/subliminally vs. consciously/supraliminally, or whenparticipants were asked to perform an indirect task rather than a directone; (2) when participants were between the ages of three and five; and(3) when participants were anxious, as anxiety is often an indicator ofunconscious conflict and impending symptom formation (Freud 1926).

In an initial pilot study (Brakel et al. 2000), we simply hypothe-sized that (1) similarity assessments could be performed unconsciously,

L i n d a A . W. B r a k e l

1140

FIGURE 3. The Entire Item Set: 4 Lists

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

and (2) that those similarity assessments performed on subliminal(i.e., operationally unconscious) items would be organized in a pri-mary process way. In this pilot experiment, using a tachistoscope, wepresented a set of six items (different, and not as well designed as thesets used in the three experiments to be discussed below) to 41 partici-pants. These six items were presented eight times supraliminally in fullconsciousness at 40 ms for 48 card presentations, and eight times sub-liminally at the objective threshold at 1 ms for 48 card presentations.We found that for similarity assessments made subliminally the resultwas significantly in the attributional direction (t (40) = 2.51, p < .016);but we also found that the result was due essentially to two items. Afterwe redesigned the items as per Figure 3, the study we call MereCatwas begun (“Cat” for categorization, “Mere” for the role that the“mere exposure” effect plays).

MereCat

The MereCat study (see Brakel and Shevrin 2003a) consisted oftwo experiments. The first was designed as a straight replication of thepilot experiment with the improved items. We presented 58 participants48 subliminal and 48 supraliminal card presentations. The 48 presenta-tions consisted of the six items of one of the lists shown (see Figure 3)delivered eight times in individualized random orders. Half of the par-ticipants got the subliminal presentations first, and half the supraliminal.In each trial, participants were asked to choose which figure was moresimilar to the top central one—even if they saw nothing at all. Theresults showed the attributional shift we predicted. Using the McNemartest—a nonparametric statistical tool well suited to test for significanceof differences between classes of responses from single subjects—wecompared the number of our participants who supraliminally chose apredominance of relational choices but subliminally showed attribu-tional predominance (13) to their opposites who showed attributionalpredominance supraliminally but relational predominance subliminally(3). This highly significant result (chi-square = 5.06, p < .03) indicatedthat an attributional shift was indeed going on unconsciously.

The second part of MereCat was designed to bring out the attribu-tional similarity response (tracking a formal quality of primary processmentation) by contrasting supraliminal/subliminal and explicit/implicitsimilarity assessment. To place the experiment within more traditionalcognitive psychology conventions, a basic design change was made.

1141

No longer were all three figures of an item presented on a single cardsubliminally. Instead the master figure was now on a separate card andpresented either subliminally or supraliminally, and the two choiceswere on a second card, always presented supraliminally followingthe presentation of the master figure card. There were four between-participant groups: (1) Ps receiving subliminal master figures and thedirect similarity task, “Choose which of the two choices is more simi-lar to that not-seen item”; (2) Ps receiving subliminal master figuresand the indirect similarity task, “Note the not-seen-item, then choosewhich of the two choices you prefer”; (3) Ps getting supraliminalmaster figure and direct similarity choice; (4) Ps getting supraliminalmaster figure and indirect similarity choice.

The rationale for using preference to indicate indirect similaritycomes from the “mere exposure” effect. In the paradigm mere expo-sure experiment, subliminally presented items are preferred over non-presented items, even though they are not recognized as having beenpresented over nonpresented items (see Kunst-Wilson and Zanjonc1980). Since Gordon and Holyoak (1983) found that the mere exposurepreference effect carries over for items that are similar to items merelyexposed, we reasoned that when participants were merely exposed tomaster figures and given two different types of similarity choices, theywould prefer the one they implicitly (indirectly) regarded as more similar.

We had three hypotheses and three sets of results. First, we hypoth-esized that both subliminal duration and indirect task would shiftresponses in an attributional direction. Hence we predicted that ananalysis of variance (ANOVA)—a statistical measure used to test forsignif icance involving an interaction of more than one factor, as wellas for significance of single factors—would reveal a significant inter-action between duration (supraliminal vs. subliminal) and task (directvs. indirect). This was indeed the case, as the interaction between dura-tion and task was significant (F (1,112) = 3.46, p < .033, one-tail).

Second, we hypothesized that when looking more specificallywithin each duration, supraliminally the explicit similarity conditionwould show significantly more relational responses than the implicitpreference condition, while the subliminally implicit preference condi-tion would be significantly more attributional compared with theexplicit similarity condition. This prediction was also borne out. Thesupraliminal mean for the explicit similarity was 15.12. (The positivenumber indicates relational predominance. If all the subjects chose only

L i n d a A . W. B r a k e l

1142

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

the relational items, the mean would be 24.00.) The supraliminal meanfor the implicit preference condition was -1.69. (The negative numberindicates attributional predominance.) These are significantly different(F (1,56) = 7.58, p < .008). In the subliminal duration, the mean forexplicit similarity was 1.88, while the mean for implicit preference was-2.69. These are also significantly different. (F (1,56) = -3.39, p <.035, one-tail). These within-duration results are summarized below:

Supraliminal Duration (F (1,56) = 7.58, p < .008)Explicit Similarity Mean 15.12Implicit Preference Mean -1.69

Subliminal Duration (F (1,56) = -3.39, p < .035, one-tail)Explicit Similarity Mean 1.88 Implicit Preference Mean -2.69

Finally, for our third hypothesis we made a nonparametric predic-tion for the four conditions in descending order of number of relationalresponses. We predicted that supraliminal explicit would be the highestby far, then subliminal explicit, followed by supraliminal implicit, andfinally subliminal implicit. And indeed, as the means show, supralimi-nal explicit similarity shows the greatest relational predominance witha mean of 15.12. Next is subliminal explicit similarity with a mean of1.88. Both preference conditions show attributional predominance,with the supraliminal mean of -1.69 nonetheless indicating more rela-tional responses than the number made in the subliminal preferencecondition, whose mean was -2.69, the most attributional and thereforethe least relational of the four conditions. (The probability associatedwith predicting the correct order out of a possible 24 orders [4 x 3 x 2x 1] is .04 and significant.)

Discussion. While alternative hypotheses to account for the variouspositive findings in the two experiments MereCat comprises wouldhave to be extremely post hoc, makeshift, and therefore unconvincing,these studies do have some problems. For example, some may considerunimpressive our finding in the first experiment: the significant differ-ence favoring the number of participants shifting from supraliminal rela-tional predominance to subliminal attributional superiority over thoseshifting in the opposite direction. Likewise, in the second experimentthe relatively small size of the subliminal means can raise a questionin the reader’s mind: Are these means really different from zero (whichin this case would mean 50% of responses attributional and 50% of

1143

responses relational)? And if these means are not signif icantly dif-ferent from zero, aren’t the f indings artifacts? The answer to bothquestions is no. The statistically significant interaction (between task,explicit and implicit, and duration, subliminal and supraliminal) wasfound, and the results of the individual mean comparisons withineach duration are statistically independent of any individual meancomparisons with zero. Thus, it would have been entirely possibleto have a subliminal implicit task mean signif icantly dif ferent fromzero in the attributional direction, but not significantly more attribu-tional when compared to the mean for the subliminal explicit task. Itis the latter comparison that was significant in our study, and it is thecomparison critical to our hypothesis.

Thus, the MereCat experiment, although perhaps not in easily obvi-ous ways, does provide evidence for the first of our general hypothesesregarding the existence of primary process: that subliminal presentationand indirect task will favor attributional / primary process responses.

LifeCat

The LifeCat experiment (see Brakel, Shevrin, and Villa 2002)addressed the second general hypothesis: that even the conscious think-ing of the very young (under 7) would be predominantly attributional /primary process in nature, as the secondary process has not yet fullydeveloped. In this simple experiment 559 participants ranging in agefrom 3 years to 79 years were given a single supraliminal (and non-time-limited) presentation of six items from one of the four lists (seeFigure 3). Participants were asked to give a direct similarity judgmenton each the six items. It was predicted that the 3-to-5-year-old pre-school group would show attributional predominance at a level reach-ing statistical significance, reflecting the power of primary processorganization in that age group, whereas all our other age groups wouldshow strong and statistically significant relational / secondary processpredominance. As Figure 4 demonstrates, these predictions were borneout. Table 1 presents the signif icant levels of either attributional orrelational predominance by age group.

L i n d a A . W. B r a k e l

1144

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

1145

FIGURE 4. Participants’ Relational Choices by Age Group

Table 1. Relational (REL) vs. Attributional (ATT) Choices:Number of Participants Favoring REL Choices vs. ATT

and Percentage of Total REL Responses vs. ATT

Discussion. Although the findings in the LifeCat experiment arestraightforward, puzzling matters remain. For example, both the adultdata and the child data show that approximately one-third of partici-pants do not give the predicted predominant mode responses. First, arethe reasons the same for both preschoolers and adults? Second, whatare the reasons in both cases? Regarding the first question, I think thevery youngest children should be considered separately on this matter.Based on their comments as they completed the items, there were likelytwo populations within the one-third who responded in a relationaldirection.2 I speculate that one group had already developed sufficientlyto show the relational predominance typical of more mature thinkers,and the other group was not yet sufficiently developed to have consoli-dated attributional predominance, choosing the relational choicesbecause these shared a global resemblance to the master figures (seeFigure 3). Gentner (1988, 1989; Gentner and Ratterman 1991) proposesthat the earliest mode of categorization is based on global similarity,with attributional categorizations following at around age 3. Regardingadults, we are currently collaborating with researchers (Nnamdi Pole

L i n d a A . W. B r a k e l

1146

2Thanks to Colleen Hastie, who as an undergraduate ran the experiment andrecorded these comments.

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

at the University of Michigan and Steven Hibbard at the University ofWindsor) who are running versions of LifeCat along with many otherpsychological measures (e.g., various personality tests) attempting tofind meaningful correlations.

The LifeCat experiment, in line with the work of some nonpsycho-analytic developmental researchers, clearly provides evidence for oursecond general hypothesis, by demonstrating that even the conscioussimilarity categorizations of preschoolers show attributional / primaryprocess organization predominance.

StressCat

The StressCat experiment (see Brakel and Shevrin 2004) addressedthe third general hypothesis toward confirming the existence of the pri-mary process mode of mental organization. The hypothesis here was intwo steps. Since psychoanalytic theory predicts that unconscious con-flict yields a shift toward primary process formal organization (as wellas content), we presumed that increased state anxiety, as a frequentindicator of unconscious conflict would also produce this shift. Then,as in the other category experiments, it was hypothesized that this shiftwould be demonstrable by a shift toward attributional similarity judg-ments. The procedure and items were the same as for the adult partici-pants in LifeCat, except for two changes: (1) the participants were 120adults in waiting rooms at various outpatient facilities at the Universityof Michigan Hospitals; (2) atop each six-item set (from the same fourlists) was the question “How anxious are you as you wait, where 1 = calmand 10 = very anxious?”

Using three different statistical measures, we found evidence forour hypothesis. First, using the 410 adult participants from the LifeCatstudy as a normative relational vs. attributional sample, the number ofparticipants who would be expected, under nonstressful conditions, tobe relationally predominant was computed. This number, 69.9% ofparticipants, when compared with the 55.5 % of individuals who wererelationally predominant in the StressCat experiment, yielded a highlysignificant result (chi-square = 7.14, p < .007). The difference was evenmore significant when only the participants at anxiety level 2 or above(i.e., participants who rated themselves as having at least minimal anxi-ety) were considered. For this group only 46.6% showed relationalpredominance. Comparing the normative sample 69.9% expected to berelationally predominant to this 46.6% yields a chi-square = 10.55,

1147

p < .001. In contrast, the group that rated themselves calm (anxietylevel 1) showed no significant difference from the normative sample, with 69.9% favoring relational choices in both cases. (These “calm”waiting room participants did differ significantly from those rating them-selves 2 or above on anxiety: chi-square = 4.01, p < .04.) The secondmeasure was a correlation of score (relational choices minus attributionalones) with anxiety measure (1 through 10). Controlling for age, educationlevel, gender, handedness, and list, we found that attributional choicescorrelated positively and signif icantly with anxiety level (partial cor-relation coefficient = .228, p < .015). Finally, a linear regression analysis(see Figure 5) showing the fall of relational responses with increase inanxiety was performed. This was highly signif icant (F = 12.95,signif icance of F < .0005).

FIGURE 5. SCATTERPLOT OF STRESSCAT RESULTS

Discussion. The StressCat experiment provides evidence for ourthird hypothesis, regarding the existence of primary process organiza-tion tracked via a predicted shift toward attributional categorization.With even mild state anxiety, there is a significant shift toward attribu-tional responses. An alternative hypothesis for these findings has beenraised by Goldstone (personal communication; Goldstone and Medin1994). Suppose that the anxiety finding was really a speed finding. Inother words, what if the anxious participants rushed their responsesand/or did not attend well to the stimuli, etc. If that were the case,

L i n d a A . W. B r a k e l

1148

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

then the role of anxiety would be just contingent and not causal, withspeed rather than anxiety mediating the shift toward attributionalresponses. We (Brakel and Shevrin) are currently designing experimentsto explore this matter. In the new studies there will be four between-participant conditions: anxiety plus speed constraints, anxiety with nospeed constraints, no anxiety but speed constraints, and neither anxietynor speed constraints. We hypothesize that anxiety independent ofspeed will tend to shift responses toward the attributional mode,although speed will likely enhance this effect.

Conclusions from the Categorization Experiments

Independent evidence for the primary process assumption hasbeen provided. The three category studies showed a shift toward at-tributional organization—the extraanalytic experimental/descriptivemode we used to index a formal aspect of primary process mentation—inprecisely the three areas indicated by psychoanalytic theory: (1) underunconscious rather than conscious conditions and with indirect ratherthan direct tasks; (2) in ordinary conscious mental operations of thevery young; and (3) when anxiety, often an indicator of unconsciousconflict, is present. Each of these three findings itself strengthens thepsychoanalytic assumption of primary process. Further, when thesethree category studies are looked at together, the psychoanalytic theoryof which primary and secondary processes are a part gains very valu-able support by consiliance. To our knowledge, no other cognitivetheory could possibly predict or explain an increase in attributionallyorganized categorization in just these disparate areas. Even the at-tributional/relational distinction in categorization, while capable ofdescriptively accounting for our findings, cannot explain them. Thesedescriptive categorization theories would have no basis on which toinvestigate the areas that were studied, much less to hypothesize andpredict the findings. I will have more to say about evidence gained froma variety of sources at the end of the article.

AN IMPORTANT AREA OF DISAGREEMENTAMONG PSYCHOANALYSTS

Many psychoanalysts have raised the criticism that the attributional cat-egorization / relational categorization distinction does not at all maponto primary and secondary process. These analysts sharply differ with

1149

us regarding the essential nature of primary process mentation andhence contest our view that attributional similarity can index an aspectof primary process organization.3 Really there are two groups of ana-lysts who disagree with us. For one group, primary process is primaryprocess content. This content is drive-infused, primitive, and pathologi-cal and clearly represents conflictual material. We have no quarrel withthis view of primary process content, but we are dealing with the for-mal aspects of primary process thought—condensation and especiallydisplacement, and the sorts of mentation that result. This clarificationnotwithstanding, another group of analysts holds that we are not deal-ing with primary process. For these analysts, primary process organiza-tion is associated with, perhaps even caused by, pathology and primi-tive regressive conflicts—all quite absent in our experiments.

Now, this disagreement is not new. Forty years ago Merton Gill(1963,1967) wondered if primary process mechanisms arise only inresponse to conflict, “to evade the censorship . . . [or if they are]present from the beginning, sui generis, as it were” (1967, p. 261). Wehave held in the past that our positive findings, even with such neutralstimuli, argue in favor of the primary process being ever present,available to be appropriated when there are unconscious conflicts,and against the idea that primary process mentation is caused by un-conscious conflicts (Brakel et al. 2000, p. 558). However, this is not atelling argument if there is indeed no validity to the initial mapping ofattributional and relational categorizations onto the primary and sec-ondary processes. We have indeed assumed this mapping and thengone on to perform the experiments. But what does strongly favor ourview that attributional similarity can index a formal aspect of primaryprocess mentation is that we found signif icant increases in attri-butional categorization exactly where psychoanalytic theory predictsthere would be significant shifts toward the primary process mode.Again, to our knowledge there is nothing in the cognitive psychologyunderstanding of attributional categorization that would even look

L i n d a A . W. B r a k e l

1150

3Robert Holt’s significant body of work investigating primary process (e.g., Holt1967, 1989, 2002) is not subject to this criticism. In line with the view presented here,Holt (2002) states, “At present, the primary and secondary processes remain hypo-thetical, inferred happenings, and we have no way to observe them directly”; hecontinues, however, reaching a conclusion very different from ours: “We mustinstead content ourselves with thought products [i.e. contents] . . . ” (p. 461). Holt’swork then, unlike that presented here, does not concern the formal organizationalaspect of primary process without the simultaneous presence of the conflictual,drive-dominated primary process content more familiar to psychoanalysts.

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

for the f indings in these various areas, much less predict andexplain them.

Our view of primary process includes its association with thepathological and primitive in the following way: pathological statesallow formal regression to primary process operations; primary processoperations are appropriated to the service of pathological/irrationalends; and primary process contents result. However, for us the primaryprocess mode as a form of mental organization is in itself neitherpathological nor even irrational. Rather, it is an early process, develop-mentally and evolutionarily, with primary process mentation oftenfunctioning effectively enough in many circumstances. We think it is themajor mode of thought operative in young humans. Further, we think itis the predominant form of thought in adults in dream states, during theformation of unconscious fantasies, and in general whenever signifi-cant mentation is operative outside of reflective consciousness (e.g., inindirect tasks such as free association). Perhaps it is the major mode ofthought organization operative in nonhuman primates, as well as inother higher mammals and some bird species. Perhaps primary processorganization underlies such basic psychological operations as assess-ments-in-action and generalization in conditioning. As I will show inthe section to follow, these speculations can be made testable. Lessspeculatively, just as primary process organization can be put in the ser-vice of pathological ends, it can serve creative ones too, as in Kris’s“regression in the service of the ego” (1935).

FUTURE DIRECTIONS

Much of the work presented here is preliminary. Using the same or verysimilar stimulus items as those in the three experiments discussedabove, there are three directions in which we would like to extend thiswork in which the attributional and relational modes track primary andsecondary process organization: (1) the role of primary process in theformation of psychological symptoms; (2) the role of primary processin nonpathological, very general psychological operations; (3) the roleof primary process in certain addict-like pathological behaviors toreveal the suspected role of primary process underlying an entire classof normal behaviors.

Concerning primary process and psychopathology, we certainlyrealize that our study of state anxiety captures only a small portion of

1151

the pathway from unconscious conflict to symptom formation. Thus,combining the research program discussed here with some of our earlierwork, we are constructing a new experiment. In the Shevrin et al. work(1992, 1996), words representing participants’ unconscious conflicts,presented subliminally and supraliminally, vs. those representing theirconscious symptoms, presented subliminally and supraliminally, pro-duced significantly different evoked-response potential (ERP) responsepatterns. (ERPs are electrical measures of brain wave activity). Usinga mathematically sophisticated time-frequency information analysiswhereby certain features of the ERPs can be used to test how well theERPs to a certain class of stimuli can themselves be classified, we foundthe following. Across participants, subliminally the time-frequencyfeatures classified the unconscious conflict words’ ERPs significantlybetter than they did the conscious symptom words’ ERPs, whereassupraliminally the reverse was true—the conscious symptom words’ERPs were classif ied by the time-frequency features signif icantlymore successfully than those to the unconscious conflict words. (Thislatter part of the finding was corroborated by another task the partici-pants performed. They were each given a list of all of their words—unconscious conflict words and conscious symptom words—and askedto put them in categories. The conscious symptom words were mostoften put in a single category, while the unconscious conflict wordswere dispersed over three or more categories.) Following on this evi-dence, the new study would be designed to assess whether these differ-ences have psychological consequences in terms of primary processvs. secondary process organization. In particular, we propose to answerthe following question. Do participants, given a simple supraliminalsimilarity judgment test (e.g., one of the lists from the LifeCat exper-iment; see Figure 3) immediately following subliminal (unconscious)exposure to their unconscious conflict words, make more attributionalchoices than they do either after supraliminal exposure of these uncon-scious conflict words, or after exposure to conscious symptom words,subliminally and supraliminally? Our hypothesis in all three caseswould be yes; if borne out, this would suggest a shift toward primaryprocess organization (indexed by an attributional categorization shift)after subliminal (unconscious) exposure to words representing a par-ticipant’s unconscious conflict.

Regarding primary process in nonpathological, very general psycho-logical operations, we speculate that primary process mental organization

L i n d a A . W. B r a k e l

1152

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

is more widespread than even Freud imagined. For example, just as evi-dence has been found that primary process is the predominant systemof categorization in preschoolers, we believe evidence can be soughtfor primary process organization dominating in other species—atleast in nonhuman primates, and perhaps in other mammals, includingdogs, cats, and rats, and some bird species.

If such evidence could be gained, this would suggest that primaryprocess organization is not a manifestation of the irrational—a patho-logical, regressive form of rational thought—but instead that the pri-mary process mode is arational and sufficient for effective functioningmuch of the time. I have attempted to make a philosophical case for thearational as truly mental (Brakel 2002), and we have tried observation-ally to demonstrate examples of the primary process system in non-human species (Brakel and Shevrin 2003b). But we need empiricalevidence. We have designed two simple experiments involving birdsor small mammals and are seeking collaborators with animal labs.4

In these experiments an animal would be classically conditionedsuch that a master figure (such as those used in the human experimentspresented; see Figure 3) would be paired either with a very pleasurablestimulus or with an aversive one. After conditioning was established,we would rely on stimulus generalization in the following way. Animalswould make a forced choice between an attributionally similar vs. a rela-tionally similar item. (The choice would be appropriate to the animal—for a pigeon, pecking on an attributionally marked key vs. a relationallymarked one; for small mammals, entering an arm of a run or pressinga bar with an attributional marking vs. one with relational marking; etc.)In the pleasurable conditioning experiment, the animal conditionedto “expect” the reward would pick the attributional or relational choice,with this choice indicating that this was the one more readily general-ized from the conditioned master figure (i.e., as the one more similar), inorder to gain the reward again. In the aversive conditioning experiment,the avoided choice is the one more readily generalizable—more similar. Itis avoided to prevent repetition of the aversive stimulus. In both experi-ments we would hypothesize that the attributional mode is more readilygeneralizable, and in this experiential sense more similar.

Note that even in the manner that I am considering the animal ex-periments, my speculation about the predominance of primary process

1153

4One such collaboration is under way in the laboratory of Aaron Blaisdell(UCLA), assisted by Seth Resnick (Mt. Sinai Medical School) with pigeon subjects.

organization in ubiquitous, nonpathological, general psychologicaloperations is revealed. To develop this line further, it can be notedthat in the proposed conditioning experiments similarity judgmentsproper are not being made; instead, some sort of similarity assessmentis being acted on. I am suggesting that perhaps such similarity-in-actionassessments are primary process mediated, whereas judgments of simi-larity are more secondary process mediated operations. And this couldbe tested empirically. Suppose the experiments above are run on humanparticipants, with master f igure conditioning both to highly plea-surable and to aversive stimuli. Then, in addition, there are two otherbetween-participants groups. These participants would not actuallyreceive conditioning to any master f igures, but would be told thefollowing: “In other experiments we have paired this f igure we arenow showing you [showing participant a master figure] with a verypleasant stimulus. Supposing you wanted to have that pleasant stimulusrepeated, which of these two choices [participants would be shown oneattributionally similar and one relationally similar f igure, of courseunlabeled] would you pick?” Similarly, the aversive stimuli group wouldbe told of the pairing between the master f igure and the aversive stim-ulus and asked which of two choices—one attributionally and one rela-tionally similar—they would choose so as not to suffer the aversivestimulus again. We would predict that the groups actually conditionedwould be performing similarity-in-action assessments and wouldthereby show a considerable shift toward attributionally mediatedresponses, compared to the groups asked to make judgments-properof similarity. It would be of considerable interest if it could be estab-lished that the judgment process itself requires secondary process organi-zation (as indexed by increased relational choices). It would be perhapsof even greater moment if, using increased attributional choices as anindex, it could be demonstrated that something as basic as the inductionof generalized conditioned responses in fact rests on primary processmental organization.

Finally, there is the work we would like to undertake concerning therole of primary process in certain pathological, addictive-like behav-iors—a role we hypothesize would reveal the underlying function of pri-mary process in a class of important normal behaviors. Shevrin (2003)cites the work of Kent Berridge (Berridge and Robinson 1995) and JaakPanksepp (1998) in support of maintaining a sharp separation betweenthe concept of drive and that of affect (see also Shevrin et al. 2004).

L i n d a A . W. B r a k e l

1154

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

Berridge’s distinction between ”wanting” and “liking” and Panksepp’sSEEKING system—clearly drive-driven and not necessarily hedoni-cally based—both provide welcome evidence for psychoanalytic drivetheorists. Now suppose we found that pigeons and small mammals couldperform the simple category studies discussed just above; and supposewe did find that these animals categorized in a predominantly attribu-tional direction. Then consider Panksepp’s finding that when the criti-cal dopaminergic brain circuits activated during “seeking” behaviorare electrically stimulated, rats and pigeons engage in two activitiesthat look very much as though they are mediated by the primary process.(The circuits involved are in the extended lateral hypothalamic corridorfrom the ventral tegmental area/VTA to the nucleus accumbens, orNA—an area Panksepp refers to as the nucleus accumbens systems,or NAS.) One activity concerns animals engaging in a variety ofconsummatory behaviors following electrical stimulation, and not justthose matching the physiological need states rewarded, but those thatare connected only contingently (i.e., in a primary process fashion)to those states. The second activity concerns “autoshaping” behaviorin which animals treat the conditioned stimulus (the one that has beenassociated with the physiological reward)—the bar press, the key orlight to peck—as though it were itself the reward (i.e., the representa-tive nature of the stimulus is lost, as is the notion of sequence, muchas in primary process mediated hallucinatory wish fulfillment). Ourwithin-subject experiment would consist in using our simple attribu-tional vs. relational categorization task to compare animals beforeand after electrical stimulation of their nucleus accumbens systems.(Similarly, we could in a between-subject experiment compare theattributional scores for nonstimulated animals with those who had NASelectrical stimulation.) We hypothesize that electrical brain stimulationof this dopaminergic system would produce a significant increase inattributional categorizations.

The electrical brain stimulation of the nucleus accumbens systemcan serve as a model for some behaviors seen in addiction. Cocaine andamphetamines stimulate the NAS and as such have, quite independentof any hedonic pleasure, primary process mediated effects analogous tothe two described above. First, drug cravings will follow on a growingnumber of cues (internal and external) only contiguously related (relatedin a primary fashion) to the drug. Second, the drug-seeking behavioritself, and not just the drug, will become part of the goal pursued. Now

1155

this is clearly a pathological role for the primary process mode ofthought. But wouldn’t a positive result also suggest that it is primaryprocess mediated mentation that accounts for some of the nonpatho-logical phenomena we see under routine drive pressure? Take, for ex-ample, using a representative of the satisfying object as the object itselfwhen it is absent. Transitional objects f it this description. Also, whena loved one is absent (temporarily or permanently), some personaltokens of the lost person may be valued over and above their merematerial status. Further, despite knowledge of its impossibility, onemight be unable to stop anticipating the absent person’s appearancein situations where that person had always been present—the humanequivalent of continuing to press the bar associated with the foodreward. A grieving friend of mine, for example, for months wouldawaken in the night sure that his late wife was again in their bed. Then,regarding the even more prevalent phenomenon of the relative inter-changeability of objects of satisfaction, this too occurs under normalconditions, with no more than average drive pressures. Psychoanalystshave of course long recognized this interchangeability as a kind ofdisplacement, but empirical work along the lines we propose might gosome distance to explaining this very useful, if so far merely descrip-tive term. That drives in fact do not have univocal, singularly determi-nate objects, but instead are satisfied by any of a class of objects linkedin a primary process fashion, is a subject of philosophical importance(see Brakel 2004; Brakel and Velleman 2003).

CONCLUSIONS

All sciences have assumptions. For a science to progress, theseassumptions must be tested using methods independent of those thatpresuppose the assumptions to be true. Using a method independentof psychoanalytic assumptions to classify modes of thinking—theattributional vs. relational similarity categorizations described by cog-nitive psychologists—and recognizing that these map onto aspects ofprimary and secondary process thought organizations, respectively, wehave been able to test one of the core assumptions of psychoanalysis.We have found evidence for a preponderance of attributionally orga-nized categorizations in the very places that Freud observed the pre-dominance of primary process mentation. Freud observed that youngchildren think in a primary process manner, and we found that the

L i n d a A . W. B r a k e l

1156

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

preschoolers in our study made similarity categorizations that wereoverwhelmingly attributionally organized. Freud noted primary processcontent in dreams and symptoms, as well as during free association, andposited that it predominates in the unconscious. We found significantshifts toward attributional organization in adults under state anxiety(often reflecting unconscious conflict), as well as under conditions out-side of conscious awareness, including situations in which the mentalcontents were totally unconscious and in which indirect tasks werecalled for (both laboratory analogs of free association). These resultsprovide independent support for the psychoanalytic assumption oftwo distinct modes of mental organization—the primary and secondaryprocesses. Moreover, to our knowledge there is not another theory ofmind that would have had grounds to look for, much less predict, thedistribution of attributionally vs. relationally organized mentation insuch diverse areas over which these three experiments range: child devel-opment, state anxiety, and contents and processes outside of awareness.

This matter of diverse evidence is of no small importance. AsHellman (1997), a philosopher of science, has noted, “One of themost signif icant virtues a good scientif ic theory can possess is acapacity to account for a wide variety of phenomena hitherto regardedas disparate” (p. 202). In an attempt to provide even more diverse evi-dence, we are emboldened (1) to extend the method of indexing primaryvs. secondary processes via attributional vs. relational choices to condi-tioned responses of nonhuman species under normal and NAS–stimu-lated conditions, and (2) to explore in humans the very process ofjudgment vs. assessments-in-action. If, as we hypothesize, attributionalorganization predominates in animals generally, and more so after NAS-stimulation, as well as in similarity-in-action assessments, we will havegrounds to regard primary process mentation as having a role (1) inthe very generalization that is intrinsic to conditioning, and (2) in drive-influenced (i.e., all) behavior. This would make a very substantial claimfor the general importance of primary process. And with this claimwould come the acknowledgment of the generativity and robustnessof this central aspect of psychoanalytic theory.

REFERENCES

BERNAT, E., BUNCE, S., & SHEVRIN, H. (2001). Event-related brain potentialsdifferentiate positive and negative mood adjectives during both supra-liminal and subliminal visual processing. International Journal of

1157

Psychophysiology 42:11–34.BERNAT, E., SHEVRIN, H., & SNODGRASS, M. (2001). Subliminal visual stimuli

evoke a P300 component. Clinical Neurophysiology 112:159–171.BERRIDGE, K., & ROBINSON, T. (1995). The mind of an addicted brain: Neural

sensitization of wanting vs. liking. Current Directions in PsychologicalScience 4:71–76.

BRAKEL, L. (2002). Phantasy and wish: A proper function account ofa-rational primary process mediated mentation. Australasian Journalof Philosophy 80:1–16.

——— (2004). Drive theory and primary process: A philosophical account.Paper presented to panel: “Psychoanalysis, Cognitive Neuroscience, andPhilosophy of Mind: Shared and Unshared Ground.” American Psycho-logical Association, (Division 39), Miami Beach, March.

——— KLEINSORGE, S., SNODGRASS, M., & SHEVRIN, H. (2000). The primaryprocess and the unconscious: Experimental evidence supporting twopsychoanalytic presuppositions. International Journal of Psycho-analysis 81:553–569.

——— & SHEVRIN, H. (2003a). Attributional and relational similarity judg-ments below the objective threshold for consciousness. Manuscriptin preparation.

——— ——— (2003b). Freud’s dual process theory and the place of thea-rational. In Behavioral and Brain Sciences 26:527–528.

——— ——— (2004). Anxiety, attributional thinking, and primary processstress. Manuscript in preparation.

——— ——— & VILLA, K. (2002). The priority of primary process catego-rization: Experimental evidence supporting a psychoanalytic develop-mental hypothesis. Journal of the American Psychoanalytic Association50:483–505.

——— & VELLEMAN, D. (2003). A philosophical account of psychoanalyticdrive theory. Manuscript in preparation.

BUNCE, S., BERNAT, E., WONG, P., & SHEVRIN, H. (1999). Further evidence forunconscious learning: Preliminary support for the conditioning offacial EMG to subliminal stimuli. Journal of Psychiatric Research33:341–347.

——— KLEINSORGE, S., VILLA, K., KUSHWAHA, R., SZOCIK, J., HENRICKS, C.,BRAKEL, L., & SHEVRIN, H. (2001). Neurophysiological evidence forunconscious memory formation under general anesthesia. Paper pre-sented at the 5th International Conference on Memory, Awareness, andConsciousness, New York, June 1–3.

FISHER, C. (1954). Dreams and perception: The role of the preconscious andprimary modes of perception in dream formation. Journal of theAmerican Psychoanalytic Association 2:389–445.

——— (1957). A study of the preliminary stages of the construction of

L i n d a A . W. B r a k e l

1158

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

dreams and images. Journal of the American Psychoanalytic Associ-ation 5:5–60.

——— (1960a). Introduction: Preconscious stimulation in dreams, associa-tions and images: Classical studies by Otto Poetzl, Rudolf Allers, &Jakob Teler. Psychological Issues, Monograph 7. New York: Inter-national Universities Press, pp. 1–40.

——— (1960b). Subliminal and supraliminal influences on dreams.American Journal of Psychiatry 116:1009–1017.

——— & PAUL I. (1959). The effect of subliminal visual stimulation onimagery and dreams: A validation study. Journal of the AmericanPsychoanalytic Association 7:35–83.

FREUD, A. (1936). The Ego and the Mechanisms of Defense. New York:International Universities Press, 1966.

FREUD, S. (1895). Project for a scientif ic psychology. Standard Edition1:295–397.

——— (1900). The interpretation of dreams. Standard Edition 4/5.——— (1901). The psychopathology of everyday life. Standard Edition

6:1–279.——— (1905). Jokes and their relation to the unconscious. Standard Edition

8:9–236.——— (1915). The unconscious. Standard Edition 14:166–215.——— (1926). Inhibitions, symptoms and anxiety. Standard Edition

20:87–174.——— (1940). An outline of psycho-analysis. Standard Edition 23:144–

207.GENTNER, D. (1988). Metaphor as structure mapping: The relational shift.

Child Development 58:47–59.——— (1989). The mechanisms of analogical learning. In Similarities and

Analogical Reasoning, ed. S. Vosniadou & A. Ortony. Cambridge:Cambridge University Press, pp. 199–241.

——— & RATTERMAN, M. (1991). Language and the career of similarity. InPerspectives on Language and Thought: Interrelations in Development,ed. S. Gelman & J. Byrnes. Cambridge: Cambridge University Press,pp. 225–277.

GILL, M. (1963). Topography and Systems in Psychoanalytic Theory. Psycho-logical Issues Monograph 10. New York: International Universities Press.

——— (1967). The primary process. In Motives and Thought: Essays inHonor of David Rapaport, ed. R. Holt. New York: International Uni-versities Press, pp. 260–298.

GOLDSTONE, R., & MEDIN, D. (1994). The time course of comparison.Journal of Experimental Psychology: Learning, Memory & Cognition20:29–50.

GORDON, P., & HOLYOAK, K. (1983). Implicit learning and generalization of

1159

the “mere exposure” effect. Journal of Personality & Social Psychology45:492–500.

HELLMAN, G. (1997). Bayes and beyond. Philosophy of Science 64:191–221.HOLT, R. (1967). The development of primary process: A structural view. In

Motives and Thought: Essays in Honor of David Rapaport, ed. R. Holt.New York: International Universities Press, pp. 345–383.

——— (1989). Freud Reappraised. New York: Guilford Press.——— (2002). Quantitative research on the primary process: Method and

findings. Journal of the American Psychoanalytic Association50:457–482.

KRIS, E. (1935). The psychology of caricature. In Psychoanalytic Explora-tions of Art. New York: International Universities Press, pp. 173–188.

KUNST-WILSON, W., & ZAJONC, R. (1980). Affective discrimination of stimulithat cannot be recognized. Science 207:557–558.

MEDIN, D., GOLDSTONE, R., & GENTNER, D. (1990). Similarity involvingattributes and relations: Judgments of similarity and differences arenot inverses. Psychological Science 1:64–69.

——— ——— ——— (1993). Respects for similarity. Psychological Review100:254–278.

MURPHY, G., & MEDIN, D. (1985). The role of theories in conceptual coher-ence. Psychological Review 92:289–316.

PANKSEPP, J. (1998). Affective Neuroscience: The Foundations of Human andAnimal Emotions. Oxford: Oxford University Press.

RAPAPORT, D. (1951a) States of consciousness. In The Collected Papers ofDavid Rapaport, ed. M. Gill. New York: Basic Books, pp. 385–404.

——— (1951b). Toward a theory of thinking. In Organization and Path-ology of Thought, ed. D. Rapaport. New York: Columbia UniversityPress, pp. 689–730.

SHEVRIN, H. (2003). The psychoanalytic theory of drive in light of recent neu-roscience findings and theories. C. Philips Wilson Memorial Lecture,Psychoanalytic Association of New York, September 15, Lenox HillHospital.

——— BOND, J., BRAKEL, L., HERTEL, R., & WILLIAMS, W. (1996). Consciousand Unconscious Processes: Psychodynamic, Cognitive, and Neuro-physiologic Convergences. New York: Guilford Press.

——— & FISHER, C. (1967). Changes in the effects of a waking subliminalstimulus as a function of dreaming and nondreaming sleep. Journal ofAbnormal Psychology 72:362–368.

——— & FRITZLER, D. (1968a). Brain response correlates of repressiveness.Psychological Reports 23:887–892.

——— ——— (1968b). Visual evoked response correlates of unconsciousmental processes. Science 161:295–298.

——— & LUBORSKY, L. (1960). The rebus technique: A method for studying

L i n d a A . W. B r a k e l

1160

THE PSYCHOANALYTIC ASSUMPTION OF THE PRIMARY PROCESS

primary process transformations of briefly exposed pictures. Journal ofNervous & Mental Disease 133:479–488.

——— PANKSEPP, J., SNODGRASS, M., CHO, J., & BRAKEL, L. (2004). In themood: A study of affect, conscious and unconscious. Manuscript inpreparation.

——— SMITH, W.H., & FRITZLER, D. (1971). Average evoked response andverbal correlates of unconscious mental processes. Psychophysiology6:149–162.

——— & STROSS, L. (1962). The effects of altered states of consciousness onprimary-process transformations of briefly exposed pictures. AmericanPsychologist 17:296.

——— WILLIAMS, W., MARSHALL, R., HERTEL, R., BOND, J., & BRAKEL, L. (1992).Event-related potential indicators of dynamic unconscious. Conscious-ness & Cognition 1:340–366.

SMITH, E., & MEDIN, D. (1981). Categories and Concepts. Cambridge: Har-vard University Press.

SNODGRASS, M., SHEVRIN, H., & KOPKA, M. (1993). The mediation of in-tentional judgments by unconscious perceptions. Consciousness &Cognition 2:169–193.

STROSS, L., & SHEVRIN, H. (1962). Differences in thought organization be-tween hypnosis and the waking state: An experimental approach.Bulletin of the Menninger Clinic 26:237–247.

WONG, P., BERNAT, E., BUNCE, S., & SHEVRIN, H. (1997). Brain indices of non-conscious associative learning. Consciousness & Cognition 6:519–544.

——— ——— SNODGRASS, M., & SHEVRIN, H. (in press). Event-related braincorrelates of associative learning without awareness. InternationalJournal of Psychophysiology.

——— SHEVRIN, H., & WILLIAMS, W. (1994). Conscious and nonconsciousprocesses: An ER index of anticipatory response in a conditioningparadigm using visually masked stimuli. Psychophysiology 31:87–101.

525 Third StreetAnn Arbor, MI 48103E-mail: brakel@umich.edu

1161