Investigating the Relative Influence of Top-Downversus Bottom-Up Processing on Viewing Ambiguous Figures

(Paper compiled from individual papers written by theseven student in PSY 440, Spring 2000)

Return to PSY 440 frontpage, Spring2000.

The influence of the human brain on visual perception is largelystudied in the field of psychology today. Understanding the waymemories, experiences and expectations are perceived can enhance theway we view perception as a whole through better understanding. Oneway of studying the way that memory and experience interact withperception is to look at the differences in true object perceptionand perceptions influenced by memory and experience. Perceptionsinfluenced by the visual field itself, or changes in stimuli arereferred to as "bottom-up" influences. If a perception is influencedby what a person expects, or has experienced before, it is referredto as "top-down" influence. One way to measure the differences inthese two types of processing is to present a stimulus to people thatcan be perceived in more than one way. Ambiguous figures are stimulithat contain two perceivable images that seem to reverse. Only oneimage in the reversible figure can be perceived at one time, allowingresearchers to examine a recordable response from participants(Stark,1999). The idea of this separation of images is commonly referred toas "figure-ground reversal", and was a major influence of Gestaltpsychologists. One of the most famous examples of this type of figureis the "face-vase" figure introduced by Danish psychologist EdgarRubin in 1915 (Goldstein, 1999). Using this type of stimulus,researchers can manipulate both the expectations of the participantprior to viewing the figure, or the actual physical characteristicsof the figure itself, biasing recognition of one image over theother.

The term top-down processing refers to the expectations, memories,biases and general upper-level influence the brain has on perception.If a person views an ambiguous figure and is predisposed to see onefigure instead of the other, then this top-down predisposition canaffect the way he sees the image. This influence has been shown to bequite strong. For example, Thomas (1999) demonstrated thatparticipants who were shown a reversible figure briefly, and thenwere asked to sketch the figure they perceived, could not identifythe other figure from their sketch. After drawing the image theyinitially perceived, participants were even told what the other imagewas, and still could not identify it in their sketch, even though thedrawings were not very dissimilar from the original. This gives us anidea of how strong the top-down influence on visual processing canbe.

Another example of this influence can be seen in a study thatdemonstrated the effect of logical motion on these figureidentifications (McBeath, Morikawa, and Kaiser, 1992). Participantswere shown reversible figures in which both figures could beassociated with motion. When the image was shown while moving in aparticular direction, participants were more likely to identify thefigure that would logically move in that direction. For instance, ifthe figures in the image were two animals, participants would be morelikely to recognize the animal facing the direction of motion.Bernstein and Cooper (1997)further demonstrated this effect in astudy showing the same bias when reversible figures were moved in alogical direction. Both of these studies show a strong influence oftop-down processing on figure recognition. The importance of top-downprocessing on figure recognition was further supported in a report byLong and Toppino (1994). In their report, they simply compiled theresults of several studies and explained how the data supported theimportance of top-down processing on visual perception.

The role of bottom-up processing can also have a strong effect onwhat people perceive. The stimulus itself can be altered in such away that it is more likely to be seen as one figure than another.Even though both objects can be recognized, perception of one can bestrengthened through altering its characteristics. One example ofthis is the ideas presented on an Internet site from theIllusionworks Company. The Internet site points out that when viewingthese reversible figures, focusing on one particular segment of thefigure can cause the viewer to be biased towards seeing one imageinstead of the other. If the focused area is more similar to acharacteristic in only one of the objects, they will most likely seethat particular object(IllusionWorks, L.L.C., 1997).

Long and Olszweski (1999) conducted two experiments to examinebottom-up processing in the reversibility of reversible figures. Theyhypothesized that numerous stimulus and procedural variables can beexpected to interact effectively in any research effort, given thecombined roles of both sensory and cognitive processes in phenomenalreversals. Stimulus control was found to be critically important inthis experiment. The size of the stimulus, stimulus location, viewingduration, fixation control, and other bottom-up variables affectedthe number of reported reversals in this experiment. Although thepurpose of their experiment was to examine the influence of bottom-upprocessing on reversible figures, the researchers noted that theobservers' prior experience with the figures, their expectations inthe setting, their attention, and other top-down variables may havealso influenced the reported reversals. This study was furtherreinforced when analyzed by Peterson (1999). He agreed that bothbottom-up and top-down processing influence perception.

The literature reviewed for this study demonstrates that aperson's perception of an ambiguous figure can be influenced inseveral ways. Through manipulating either the stimulus itself, or theparticipant's expectation of what they will see, we can expect agreater likelihood of the emphasized figure to be recognized first.In the current study, the roles of top-down and bottom-up processingon visual perception were examined. The question addressed was thedegree to which we can influence visual perception, throughmanipulations of the visual field or of the expectations andexperience held in memory.

It was hypothesized that the presentation of bottom-up cues addedto an ambiguous figure would be more influential than thepresentation of top-down cues presented before an ambiguous figure.This hypothesis was derived from the idea that the early, or first,visual processes behave mostly in a bottom-up fashion, separate fromcognitive or top-down processes(IllusionWorks, L.L.C., 1997). It was also hypothesized that the gender ofthe participant would affect the influence of the priming andperception of the figure.



Eighty-four, 42 male and 42 female, students enrolled at StephenF. Austin State University were asked to participate in this study.The majority of the participants were age eighteen throughtwenty-five. Participants were individually solicited by the sevenmembers of the research group, and received no compensation.Participants had 20/20 or corrected to 20/20 vision and had normalcolor vision. Participants were treated in accordance with the"Ethical Principles of Psychologists and Code of Conduct" (AmericanPsychological Association, 1992).


A simple, mixed design was used, with the independent variable(prime type) having seven levels (label, word scenario, picturescenario, neutral, color, detail, and lightness/ contour), and thedependent variable (participant response) having one level. Genderwas also included as an independent variable.


There were seven different reversible figures that weremanipulated in seven different ways (prime type) for each of the twointerpretations of each reversible figure. Thus, there were 49 figuremanipulations, times two. The three top-down prime type conditionsconsisted of a prime sheet containing a label, a word-scenario, or apicture. The three bottom-up manipulations consisted of a coloraddition to the ambiguous figure, added detail, or a change in thelightness contour. For example, the ambiguos figure presented inFigure 1 can be identified as either a man playing the saxophone or awoman's face.

Figure 1

For the label priming of this figure, the phrase "saxophoneplayer" appeared on the prime sheet to influence the perception ofthe saxophone player, or the word "woman" appeared for priming of thewoman. For the word-scenario conditions, the descriptions were of ajazz club or a woman's role in a film. The last top-down influencewas the picture scenario, which consisted of a picture of a trumpetor a picture of three women. Color manipulations included a goldcolored saxophone for the priming of the saxophone player, and pinkcolored lips in order to prime the woman's face. The detailmanipulation consisted of the appearance of musical notes at the endof the saxophone or chin continuation and eyelashes on the woman. Thelast bottom-up influence consisted of manipulation of the colorcontrast of the figure in which the saxophone player appearedlighter, or the woman's hair appeared darker. The packet alsoconsisted of one neutral ambiguous figure that was not exposed to anytype of priming.

In order to prevent a participant from viewing an ambiguous figuremore than once, each of the seven experimenters had a two-part packetcontaining consent and debriefing forms, and seven differentreversible figures that were manipulated in seven different ways. Thetwo parts separated the two interpretations of the ambiguous figures.For example, one part would have the color prime for the saxophoneinterpretation and the other would include the color prime for thewoman's face interpretation. Thus, each participant would be exposedto all seven ambiguous figures, and all seven prime types, but wouldnot see any of them more than once. The order of presentation of theseven levels of prime type were partially counterbalanced across theexperimenters and was then fixed for each packet.


The seven experimenters each solicited 12 volunteers toparticipate in the experiment, three males and three females for bothparts of their packet of stimuli. Before receiving the consent form,the experimenter explained to the participants that participation wasvoluntary and could be terminated upon their decision at any time.All participants were tested individually in a comfortable pace.Participants were given approximately 30 seconds to examine thecondition relevant to the ambiguous figure that was presented.Following presentation of each condition, the participants were askedwhat figure they perceived first, and this was recorded. Aftercompletion of the packet, the debriefing form was given to theparticipants.

Following the collection of data, each of the subjects' responseswere changed to a "yes" if the subject saw what was primed, and to a"no" if the subject did not see what was primed.


The results were first compiled into a table. Women's responseswere separated from the men's, and rows were made up of each possiblefigure, totaling fourteen rows per gender. Columns were made up ofthe type of prime used, totaling six columns. A final column in whichthe unaltered data were recorded was added for comparison, but wasnot used in the analyses.

Responses were recorded using percentages. For example, for thedevil word prime, one out of three women saw the primed "devil",therefore, a percentage of 33 was recorded on the table. The datawere then analyzed using a two-factor mixed design ANOVA. The betweenfactor was gender, and the type of prime that was presented was thewithin factor.

The effect of gender was not statistically significant, F (1, 26)= .91, p > .05. However, a significant effect for prime type wasrevealed, F (5, 130) = 3.94, p < .01. A Post Hoc Tukey HSD wasperformed on prime type and revealed significance pertaining to thetype of manipulation presented. The detail manipulation of theambiguous figure was significantly more influential than any of thetop-down manipulations. The color and contour manipulations were notsignificantly more influential than the top-down manipulations.

Although there were no significant results pertaining to thegender variable nor a significant interaction between prime type andgender, females tended to be influenced more strongly by the primesthan males. Specifically, the biggest difference between males andfemales occurred using the picture prime. The graph presented inFigure 2 represents the means of the male and female percentages ofresponses matching the prime type presented.

Figure 2:


Finally, when prime type was collapsed to two levels (Bottom-up v.Top-down), and the ANOVA was repeated, there was a significantdifference between the two types of priming, with bottom-up primingbeing more influential than top-down priming, F(1,26) = 7.51,p<.01.


This research was an attempt to determine whether the presentationof top-down cues presented before an ambiguous figure, and thepresentation of bottom-up cues added to the figure, would influencean individual's first perception of the figure. The topic of whetherthe gender of the person being primed influenced the perception ofthe figure was also addressed.

Tuccio (1995) stated that there are many stimulus conditions andfactors that influence an ambiguous figure's perceptual organization.Manipulating ambiguous figures by using a word label, word scenario,or picture scenario prime sheet for top-down cues, and adding color,detail, or changing the color contour for bottom-up cues was themethod used in this research in order to provide each participantwith the six different prime types. The Two-Factor, Mixed ANOVAs thatwere used in this study revealed a significant difference between theinfluence of top-down cues and bottom-up cues. In general, thepresentation of bottom-up cues was significantly more influentialthan the presentation of top-down cues. More specifically, thebottom-up manipulation that consisted of adding detail to theambiguous figure appeared to influence the participant more than anyof the top-down cues. These results may suggest that what anindividual ultimately perceives is based on certain aspects of thestimulus, such as its color, any detail that it contains, and thecolor contour of the object. This conclusion corresponds to theconclusions drawn by Cave and Kim (1999) which revealed that a colordistractor, a bottom-up cue, could draw attention to a stimulus inthe early stages of visual processing. A second possible explanationfor these results is that adding detail to a picture simply allows aperson to recognize it more effectively than color or shading, andour experiences and memories, while influencing us somewhat, cannotcompete with a real stimulus in the environment.

According to the results addressing the hypothesis that genderaffects the influence of priming of an ambiguous figure, overallthere was not a significant difference between males and females intheir responses. However, when looking at the means of the responses,females tended to be influenced more strongly by the top-down primesthan males. The biggest difference of influence occurred in the primeusing the picture scenario.

An additional suggestion that there may be gender differences eventhough the above results were insignificant comes from an examinationof the actual responses when subjects were reporting what image theyfirst perceived in the reversible figure. More specifically, for somereversible figures, the male participants tended to see differentinterpretations than those determined by the experimenters (who werepredominantly female). For instance, with an ambiguous figure thatcould be seen as either "arrows" or the word "win" some of the maleparticipants identified the arrows as "soldiers". Thus, more malesmight have matched the primed interpretation if the word or scenarioschosen had been congruent with the "soldier" interpretation ratherthan the "arrow" interpretation.

In summary, both top-down and bottom-up processing appear to havea complicated effect on the way we perceive the world. The detailspresent in an object seem to be very important to our recognition ofthat object, and females may be more flexible in their perceptions ofthe world, allowing experience and stimulus characteristics to play alarger role in object recognition. This study has not definitivelyanswered the questions it sought out to, but has perhaps raised a fewquestions in the minds of future researchers. The effects hinted atin the results of this study should be pursued on a broader level,using many more participants as well as many more reversible figuresto obtain a better, clearer picture of the effects of our memories,experiences, and the details of the environment on visual perception.



Bernstein, L. J. & Cooper, L. A. (1997). Direction of motioninfluences perceptual identification of ambiguous figures. Journal ofExperimental Psychology: Human Perception and Performance, 23, 721.

Cave, K. R., & Kim, M. (1999). Top-down and bottom-upattentional control: On the nature of interference from a salientdistractor. Perception & Psychophysics, 61, 1009-1023.

Goldstein, E. B. (1999). Sensation and Perception. (Rev. ed.).Pacific Grove, CA: Brooks/Cole.

Illusionworks, L. L. C. (1997). Perceptual ambiguity.

Long, G. & Olszweski, D. (1999). To reverse or not to reverse:When is an ambiguous figure not ambiguous? American Journal ofPsychology, 112, 41-56.

Long, G. M. & Toppino, T. C. (1994). Adaptation effects andreversible figures: A comment on Horlitz and O'Leary. Perception andPsychophysics, 56, 605-610.

McBeath, M. K., Morikawa, K. & Kaiser, M. K. (1992).Perceptual bias for forward facing motion. Psychological Science, 3,6.

Peterson, M. A. (1999). What's it a stage name? Comment on Veceraand O'Reilly. Journal of Experimental Psychology: Human Perceptionand Performance, 25, 276.

Stark, T. I. (February, 1999). Why are Experts Skeptical?

Thomas, N. J. T. (1999). Are theories of imagery theories ofimagination? An active perception approach to conscious mentalcontent. Cognitive Science, 23, 207-245.

Tuccio, M. T. (1995). Figure-ground organization in differentphases of the perceptual alternation phenomenon. Perceptual and MotorSkills, 81, 1043-1058.

Vecera, S. P. & O'Reilly, R. C. (1998). Figure-groundorganization and object recognition processes: an interactiveaccount. Journal of Experimental Psychology: Human Perception andPerformance, 24, 441-453.