David Navon (1994) Paradoxical Effects and Occam's Razor. Psycoloquy: 5(41) Paradoxical Cognition (3)

Volume: 5 (next, prev) Issue: 41 (next, prev) Article: 3 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 5(41): Paradoxical Effects and Occam's Razor

Reply to Wegner on Paradoxical Cognition

David Navon
Department of Psychology
The University of Haifa
Haifa, Israel



For four reasons, I do not find Wegner's theory a coherent explanation of paradoxical effects: (1) the theory is less general than he claims it is; (2) what it purports to explain can be alternatively explained in terms of general, widely accepted concepts; (3) it makes an assumption that is functionally implausible; and (4) its distinctive assumption has not been empirically substantiated.


attention, automatic processes, consciousness, controlled processing, incidental learning, motor set, pain, panic attacks, paradoxical effects, positive feedback, psychosomatic disorders, recall failures
1. Quite a while ago, I wrote in a paper about visual perception: "Asking a subject to ignore [stimulus elements] may be as effective as asking him not to think of pink elephants" (Navon, 1981, p. 8). A slight modification of this sentence seems suitable as a motto for this reply to Wegner's (1994) commentary: "Asking a person not to see what he is driven to see may be as effective as asking him not to think of pink elephants." I will try nevertheless.

2. Wegner claims that his explanation of paradoxical effects is quite general. If it could account for all, or the larger portion of, the phenomena I discuss, it would indeed be a broad theory, but I doubt, that it could do so. It attributes paradoxical effects to the failure of the intentional operating process to outweigh or suppress the output of the automatic monitoring process under conditions of REDUCED CAPACITY. How, then, can it account for enhanced failures of recall following an EFFORT to recall? Or for the disruptive effect of ATTENTION in certain perceptual and learning tasks? Or for the harmful effect of ATTENDING to habitual functions? The root of the problem in these cases must be attention or the controlled process itself. As to parsimony, James (1981/1890, p. 428) wrote: "Occam's razor, though a very good rule of method, is certainly no law of nature." When one explanation is not enough to account for a range of related phenomena, it does not seem too wasteful to posit two. The two do, however, share something: a voluntary attempt to violate the division of labor between controlled and automatic processing.

3. Parsimony is desirable, though, when a few concepts can do the same job as many, especially when the few are widely accepted for good reasons. In this respect, Wegner's assumption that evasion failures are due to a monitoring process may be unnecessary. The initial attempt to evade creates an activation, which -- coupled with priming from the mental context and a temporary lack of suppression -- gives rise to the failure. I may be wrong, but more data are required to show I am.

4. Furthermore, processes subserve functions, and they probably do so in a reasonable way. There is nothing wrong with Wegner's operating process, but his monitoring process is inherently of no avail. To evade thoughts about a concept requires that consciousness be unaffected by the activation of that concept in memory. This can, in principle, be accomplished through displacement by other thoughts, through inhibition, or both. Both types of evasions may conceivably be effected regardless of actual activation (e.g., inhibit any signals from the white bear node until the experiment is over). But suppose their initiation does require preattentive detection of activation. Four claims about the detection of activation must be valid:

    (1) Detection does not require searching all of memory; checking
    activation in the relevant node is sufficient.

    (2) It is clear that a concept can emerge in consciousness even if
    activation is not detected; thus detection is required for evading
    it, not for being aware of it.

    (3) Detection need not increase the likelihood that the
    to-be-evaded concept emerges in consciousness: the communication
    between the detecting process and the evading process can be
    designed to be inaccessible to consciousness.

   (4) A functional system would be likely to evolve in this way.

Wegner's monitoring process, however, is assumed to increase both the activation of the to-be-evaded concept (a detector that amplifies what it detects!) and the likelihood of having it in consciousness, thereby making life harder for the operating process. What is it good for, then, other than serving as the deus ex machina in Wegner's theory? The fact that unwanted thoughts are often successfully evaded clearly does not show that the monitoring process does a fine job most of the time. That may happen in spite of it, or worse yet, in spite of its nonexistence.

5. Has Wegner's theory been empirically supported? He has collected a great deal of data, but most of it just demonstrates the dependence of paradoxical effects on load. This finding, however important, can be accounted for by almost any theory I can think of, including the explanation I suggest (see para. 15 in target article, or para. 3 above). Selectivity requires effort, hence load impairs selectivity. As I have worried elsewhere (see Navon 1984, p. 232), some people think that load effects are sufficient to support a theory that assumes capacity interference, among other things. What should be tested, however, are those "other things." The distinctive assumption in Wegner's theory, namely, the existence of a monitoring process, let alone the properties attributed to it, would require different sorts of data to be substantiated.

6. In parts of his commentary, Wegner reminds me of a manufacturer reacting to the loss of his monopoly. Since the issue has been brought up, let me be explicit: nobody can get a patent for an idea that has probably been around for centuries. Trying not to think of pink elephants is an English idiom, so I'm told. The elusiveness of thought control is a central theme in oriental doctrines like Zen (Sekida, 1975). Human difficulty in abiding by seemingly arbitrary restrictions (e.g., the forbidden box of Pandora) or silly ones (e.g., the husband who orders his wife to refrain from thinking) is a recurrent motif in folklore (cf. Thompson, 1966). One example is the folk tale about the gullible Bedouine (target article, first epigraph); I had read it as a child, in an Israeli children's magazine (Toporovsky, 1949). Apparently, its impact was long lasting: I have addressed the phenomenon twice (Navon, 1981, 1994). The image of the pink elephant trampling the white bear accordingly looks somewhat -- how shall I put it -- imaginary.


#1. After locating the source of my childhood memory, I found out that the trickster was actually a peddler (not a dervish), the stuff to be transformed to gold was herbs (not straw), and the taboo thought was, believe it or not, a white bear (not a pink elephant). So, in folklore bears win after all.


James, W. (1981) The principles of psychology (Vol. 1). Cambridge, Mass.: Harvard University Press. (Originally published in 1890).

Navon, D. (1981) The forest revisited: More on global precedence. Psychological Research, 43, 1-32.

Navon, D. (1984) Resources - a theoretical soup stone? Psychological Review, 91, 216-234.

Navon, D. (1994) From Pink Elephants to Psychosomatic Disorders: Paradoxical Effects in Cognition. PSYCOLOQUY 5(36) paradoxical-cognition.1.navon.

Sekida, K. (1975) Zen Training: Method and Philosophy. New York: Weatherhill.

Thompson, S. (1966) Motif-index of folk-literature (2nd ed.). Bloomington, Indiana: Indiana University Press.

Toporovsky, Y. (1949) Once there was a Bedouine. Davar Le'yeladim, 20, 4, 68 (In Hebrew).

Wegner, D.M. (1994) Pink Elephant Tramples White Bear: The Evasion of Suppression. PSYCOLOQUY 5(40) paradoxical-cogniton.2.wegner.

Volume: 5 (next, prev) Issue: 41 (next, prev) Article: 3 (next prev first) Alternate versions: ASCII Summary