Food-Induced Attention Deficit
Hyperactivity Disorder: The Research

Home ||| Research Menu Page

Shula S. Edelkind, 1998

Attention Deficit Hyperactivity Disorder (ADHD) is a troublesome behavioral malady affecting perhaps 5% of American children. (Holmes, 1997, p. 405). Despite batteries of tests to measure attention, the actual cause of ADHD is not yet understood. Diagnosis is still based largely on hearsay and observation. After years of research, ADHD remains an enigma: a disorder without a known cause, without a definitive diagnostic test, and without a cure.

Since the 1930's, treatment has included stimulant drugs, with Ritalin (methylphenidate) as the current treatment of choice (Armstrong, 1995, p. 7). Some parents resist using such medications, while others become disturbed when their children experience side effects. These parents may become part of the large market for "alternative" therapies and products. Some alternative therapies may be useless or even harmful. Others are supported by years of scientific research but nevertheless are ignored or ridiculed by allopathic physicians.

One of the long standing alternative therapies for ADHD is the Feingold diet. Although used for almost 30 years by parents, most physicians do not offer it as a treatment. Some state medical boards consider use of the Feingold diet an "extreme departure from the Standard of Practice," for which a physician could lose his or her license. (Medical Board, 1998). And the research? The research seems to fall into opposing camps, pro and con, both types published in peer reviewed journals.

Feingold diet description.

Basically, the Feingold diet excludes synthetic (petroleum-based) dyes and flavorings, the petroleum-based preservatives BHT, BHA, TBHQ, and salicylate-containing products. This excludes several temperate-zone fruits, but most tropical fruits are acceptable. The only vegetables excluded are tomatoes, cucumbers, and peppers. Medications and products containing salicylate or aspirin are avoided, as are synthetic fragrances. The Product Information Center (PIC) of the Feingold Association of the United States (FAUS) provides regional Foodlist books listing acceptable products, updated monthly (Feingold Assoc, 1998). A number of other additives, such as MSG or sulfites, are noted in the Foodlists for those who must avoid them or wish to avoid them. Parents are encouraged to keep a daily diet diary with behavioral notes when beginning the diet. The FAUS Help-Line is a free service for trouble-shooting and helping members individualize their diet (Feingold Assoc, 1998)

Feingold diet research protocol.

For the sake of consistency in research, below is a list of parameters that I believe should ideally be followed. As I review the studies, the notation "[#__]" will refer to a violation of the numbered item in the list below. Each is a link which, when clicked while on the Internet, will bring the appropriate note into the window at the right. If reading this off-line, refer to the numbered notes below.

  1. The clinical diagnosis of participants should be appropriate. A study designed to research the connection between ADHD and diet should use participants who actually have ADHD and not something else. It may be worthwhile doing research on the connection between irritability or depression or asthma and diet, but participants must be diagnosed accordingly.

  2. Tests should be defined and agreed upon which measure not only attention and activity, but the actual symptoms such as irritability, tantrums, and sleep habits that parents observe as improved.

  3. Observation or recording of the child's behavior by parents or teachers should be done daily, at least, since daily recording is a much more sensitive barometer than weekly recording.

  4. Group data averages are fine as far as they go, but individual data must be presented and compared since it has been postulated many times that only a small number of ADHD children "really" respond to dietary intervention. Therefore, each child should be compared in the different states to himself, and the responders should also be compared to the nonresponders, not averaged in with them. If you want to weigh the needles in a haystack, you must first find a magnet that can pull the needles out of the stack, rather than weigh the whole stack and divide by number of straws+needles.

  5. Since the Feingold hypothesis involves the elimination of multiple items, then the multiple items need to be eliminated to test it. The Feingold membership materials and Foodlists should be used since the test would then be of the Feingold diet as it currently exists. Testing a 3-food diet is a valid test of a 3-food diet, but it is not a test of the Feingold diet.

  6. Since the Feingold hypothesis involves the elimination of multiple items, then the multiple items need to be used in the challenge materials to try to provoke a behavioral response in the double blind portions of the study. If a single-item challenge does not provoke symptoms, it means little, since the interactions of several items may be needed to do so.

  7. When administering challenge materials, behavior should be monitored periodically beginning one-half hour after ingestion, and ending four to six or more hours later. A single evaluation four hours after ingestion may totally miss the reaction, especially with food dyes (Adams, 1981; Swanson, 1980).

  8. When administering challenge materials, the level of such materials should be reasonably large, not to miss the "Halloween effect."

  9. When administering challenge materials, the time period of administration should be long enough to overcome what FAUS calls the "washout effect," where a child established on the diet does not react to a single infraction and can "get away with" cheating if not done frequently. Several days in a row should be adequate.

  10. Arrange to avoid cheating. Cheating twice a week may be enough to prevent improvement (has the alcoholic recovered if he only gets drunk twice a week?)

  11. Since the Feingold diet is administered and adjusted individually, some provision for doing that should be included in the study. It is not a one-size-fits-all program.

  12. The researcher must be vigilant against confounds like coloring in non-food items such as for skin care, finger paint or play dough, environmental toxins such as pesticides, fresh paint, or fragrances, etc., as outlined in the Feingold Handbooks (1982, 1996, 1998).

  13. The researcher should be aware that children recently on psychoactive medication may take up to six weeks to respond to diet therapy (Feingold Assoc, 1982, 1996, 1998). The reason for this is not known. They should either be given an extended washout period or not be used in the research.

The beginning of the Feingold diet.

In the late 1960's, Benjamin Feingold, Chief of Allergy at Kaiser-Permanente Hospital, developed the KP diet for people with allergies or asthma and aspirin sensitivity (Feingold, 1974). Because of patient reports of behavioral and personality changes on this diet, he later tried it on children who were hyperactive without allergy or asthma. About half of them reported greatly improved behavior and another 25% reported some improvement in the clinical trials (Feingold, p.71). He spent his retirement teaching the diet protocol directly to groups of parents, some of whom later founded the Feingold Association of the United States (FAUS).

Early studies.

Many of the early supportive studies were short open clinical trials, which could have involved placebo or Hawthorne effects, so they will not be discussed here. One of the early double-blind studies, by Williams, Cram, Tausig, and Webster (1978), compared stimulant medications with diet by combining active and placebo medication with two cookies containing 13 mg of mixed dyes or no dyes [#6] [#8] [#13]. The typical amount of dye ingested by a child is between 76 and 150 mg (Hersey, 1996, pg. 454; Swanson & Kinsbourne, 1980), and Williams acknowledged also that the cookies were very large and the children often did not even finish eating them (Hersey, personal communication, March 13, 1998; NIH, 1982). Williams et al. also did not eliminate salicylates, but only synthetic coloring and flavoring, a violation of [#5]. Parents kept a daily diet diary and a daily behavioral checklist, and teachers also completed behavior ratings twice a week. These ratings were the sole measurement of the children's response, yet missing ratings were filled in by estimates, and some data were dropped because since there were 26 children but only 24 possible choices of experimental sequences, two of the children were randomly assigned to an experimental sequence already "occupied." In each pair, the ones with the most missing data were simply dropped. (That must be a violation of something, but it is not on the list above). Behavioral ratings by parents at home may have suffered from the effects of the treatment wearing off by late afternoon since these children took their medicine and their cookies in the morning and at noon [#7]. The pills consumed by the 26 children were colored tablets in all four treatment sequences. When asked how he studied the Feingold diet while using pills with colorings that the diet eliminates [#12] , Williams responded that it was "only a little bit" of coloring (Hersey, personal communication, March 13, 1998). Williams et al. also documented one or two infractions (cheating) per week per child, but acknowledged that there were probably more [#10] . Seven children (27%) did well enough in all sequences to call their diagnoses into question, but their data were included in the averages [#1] . Individual results were not shown [#4] . Williams et al. reported that the children did best on medication plus the natural cookie, and worst when they were on the placebo plus the challenge cookie. The authors concluded that one fourth of the children apparently had significant diet-related behavioral change but that the children did best on stimulant medication.

Another early study, still widely quoted, was by Harley et al. (1978). It was an ambitious project funded by a large grant from the Nutrition Foundation, a food industry organization, (Hersey, 1996) in which all food was provided for the families of 46 children. Psychoactive medication was discontinued before the baseline observations. The children were randomly assigned to one of two groups: the control-diet-first group (CON-EXP) or the experimental-diet-first group (EXP-CON). Care was taken to provide diets equal in palatability, to distract parents by changing diet contents weekly, to package the control and experimental treats in identical wrappers, etc. The experimental diet was the Feingold diet, but the listing of excluded salicylates is incomplete [#5] . Toothpaste was replaced with baking soda and medication ingredients were checked, but the researchers did not address the preservatives BHA and BHT [#5] or synthetic fragrance [#12] . Preparing the experimental diet, the researchers trusted ingredient labels which is an unreliable method [#5] . Feingold had recommended contacting manufacturers, using single-ingredient items, and not relying on the word "pure" on labels (Feingold, 1974; Hersey, 1996). (FAUS Foodlists were not available at that time.) Each child received 50 mg of Vitamin C daily because oranges, a salicylate, were excluded. This vitamin added a new variable which could have been a confound, or could have been a salicylate if it contained rose hips [#12] . (Non-salicylate fruits such as lemons, pears, grapefruits, kiwi, papaya, melons, etc. (Hersey, p. 462) also contain Vitamin C.)

Average infractions (cheating) reported in the study were approximately once a week [#10] . In a letter (January 24, 1977) Harley acknowledged an average of one to two infractions per week. No correlation was made between perceived improvement and number of infractions.

Parents and teachers were asked to fill out a Conners P-TQ behavioral ratings form only once a week [#3] . Three times a week, the children were observed in school in comparison to a matched non-hyperactive control child. Such a comparison may have prevented seeing any improvement. [#4]

Nevertheless, 13 of the 36 mothers of school-age boys (36%) rated their child's behavior improved on the experimental diet, and 14 of the fathers rated them improved. Teachers did not notice any diet effect, which fit Feingold's observation that classroom behavior usually improved more slowly (Feingold, 1974). Harley et al. (1978), however, interpreted teacher reports as more reliable than parent observation, thus indicating no change. Laboratory observations of those who showed a positive response were not reported separately from those who showed no response, but were only averaged, and results were variable. Individual results were not reported [#4] . Moreover, almost all the children who improved were in the CON-EXP group (N=19). In that group, mothers reported 12 of 19 (63%) as improved on the experimental diet. While the average improvement shown by the CON-EXP group as a whole was significant, but not very dramatic, no individual data were reported [#4] . It was also not indicated whether the 14 children with neurological abnormalities were among the improved, or the un-improved. If the former, it would have corroborated an earlier study showing corrected EEG abnormalities in 9 of 20 children on an elimination diet (Kittler & Baldwin, 1970, cited in Boris & Mandel, 1994).

Harley et al. (1978) reported the CON-EXP order effect, citing a similar one in a previous study (Conners, Goyette, Southwick, et al., 1976, cited in Harley et al.). They wrote off this order effect with the comment "Whatever the reason, the fact that the experimental diet seems to "work" only when a control diet is given first would appear to attenuate the claimed efficacy of the experimental diet" (Harley, 1978, p. 826). They did not consider that the children in the CON-EXP group had been off medication three weeks longer than the EXP-CON group before being exposed to the experimental diet, which may have explained the order effect in both studies [#13].

The pre-school children were less likely to be on psychoactive medication because (a) their group was studied in the summer, and (b) such medication was then unusual under age six (Feingold, 1974; Hersey, 1996; PDR, 1982, p. 850). If not on psychoactive medication, they would not have needed a period of withdrawal before diet took effect. Indeed, the preschoolers did not exhibit any order effect, and 10 out of 10 mothers reported improvement on the experimental diet. Harley et al. explained this by saying that parental ratings would be unstable over time and unreliable without teacher corroboration (no teachers were involved); that the Conners P-TQ parent form they had used was inappropriate for such young children, and that since locomotor activity observations did not show a significant change [#2] , parent observations could be discounted. No graphs or tables of average or individual or any other kind of data were reported for the preschoolers [#4] . Thus, in the face of an unequivocal 100% of mothers in a double blind study who saw their preschool children improve on the Feingold diet, and 63% of mothers in the CON-EXP group who saw their school-age children improve, the researchers reported that the study "yielded no support for the Feingold hypothesis" (Harley et al., 1978, Abstract).

Nine years later, Gross, Tofanelli, Butzirus, and Snodgrass (1987) published a study on 39 children, of whom only 18 were hyperactive [#1] . Of those 18, all but one were on behavior modifying medications during the entire study [#12] [#13] . Although using a summer camp milieu to prevent cheating, the researchers did not address additives and colorants present in medications, toothpaste, cleaning supplies, etc. [#12] They presented an unpalatable (their description) "Feingold" diet for a single week, followed by an "additive-rich" diet the next week. For children this age, even without medication, it takes a week or more to begin to see any change when starting the diet (Feingold, 1974). Under these parameters, Gross et al. concluded that the "Feingold diet has no beneficial effect on most children with learning disorders...." and moreover that it was "distasteful to the typical American child" (Gross et al., Abstract).

Later studies.

In 1993, Carter et al. published a study in which the behavior of 59 of 78 children (75.6%) improved during an open "few foods" elimination diet. This was followed by a placebo controlled double blind challenge protocol for 19 of them that showed these children's problem behavior and psychological test performances could be reliably provoked by blind challenge. Carter et al., as opposed to Harley et al., concluded that parents are quite reliable observers.

The children for the Carter et al. study had been referred by physicians to a special diet and behavior clinic set up at the Hospital for Sick Children in London. The study is thus open to the criticism of self-selection of participants, and in fact 90% of Carter's diet responders had previously noticed a reaction to foods. On the other hand, if diet doesn't work but is only a "belief system" as Herbert (1995) calls it, then this wouldn't matter in the double blind portion.

Of the 78 children who completed the open trial phase, all but 14 had some combination of physical problems such as asthma, headaches, eczema, rashes, gastrointestinal symptoms, enuresis, or fits. Many of these physical symptoms improved on the diet, and worsened during challenges.

The Carter et al. diet was extremely limited but the foods listed were acceptable on the Feingold diet with the possible exception of the unlisted margarine ingredients. Medication, toothpaste, soap, fragrances, etc. are not mentioned [#12] , but may have been addressed since so many children responded well within the short three-week open diet period. Suspected foods and additives were then re-introduced slowly to discover which ones would provoke reactions. Food dyes were re-introduced by capsule as a dye mix maximum of 26 mg daily [#8] which, as was seen in the Williams et al. study was quite low.

During the double blind challenge period of the Carter et al. (1994) study, single suspected items [#6] and placebo were each given for one week with a two week washout period between them. Behavior was assessed at unreported intervals [#3] [#7] by the parent, using the Conners' forms, by direct psychologist observation, paired associate learning test, and the matching familiar figures test. Each child acted as his or her own control. Of the 19 children, 14 were rated as better during the placebo period. The other 5 were considered treatment failures; it is unknown whether an increase in amount or combinations of suspected provoking agents would have made a difference, because it was not tried [#8] . Again, the results were graphed for the mean, and not for individual participants [#4] , but the differences between placebo and active substance was significant, especially behavior suggesting irritability and fidgetiness rather than inattention. If irritability rather than attention is a major observation, it may explain the discrepancy between the parent reports of improvement and lack of laboratory corroboration in the Harley et al. (1978) study. (In fact, some of the laboratory tests in the Harley et al. study, such as finger-tapping speed, showed improvement during the additive-containing control diet, and one can picture the child with increased irritability tapping away.) In the Carter et al. challenge trials, the only test that did not show a significant difference was the paired associate learning test, although such a difference was demonstrated in the earlier Swanson and Kinsbourne (1980) study, in which 100 to 150 mg of mixed food dye were used rather than the 26 mg used here. [#8]

The problem of possible participant bias also plagued another study: the Boris and Mandel (1994) study which used 26 consecutive ADHD children referred to their allergy practice. Their open elimination diet prohibiting artificial colors, preservatives, and several common foods including milk, wheat, corn, yeast, soy, citrus, egg, chocolate, and peanuts, produced improvement in 73% of the children in only two weeks, followed by a dramatic response in the double blind trial, reported as graphs of averages as well as individuals. The study has been criticized because Boris is an allergist and the group of participants was small. However, if there is no diet-behavior connection, but only a "belief system," (Herbert, 1995; Medical Board, 1998) the improvement would have disappeared on the double-blind portion, even in a small sample. Also, if a strong diet-behavior connection exists within the allergic subset of ADHD children and nowhere else, that, too, is important.

Suggestions for future research.

To address the problem of self-selection for diet-responsiveness or allergy, children need to be recruited from an ADHD clinic or specialist. Children who have spent months or years on behavior modifying medications may have an altered response profile, so subject children should ideally be newly-diagnosed or at least those previously on medication should be considered separately and compared. Furthermore, anyone wishing to study the Feingold diet should administer the diet following the guidelines of the Feingold Association, not make up their own rules. Otherwise, no matter what their results, they cannot honestly say they have studied the Feingold diet. Other ideas to include in a future study or series of studies would be to:

  1. Report results for children with neurological and/or allergic co-morbidity separately to see if they respond differently to diet therapy.

  2. Report results for children by age, and plan a longitudinal study to see if children beginning diet therapy at an older age have more neurological or academic impairment.

  3. Report results for girls separately from boys.

  4. Document depression in female siblings of ADHD boys, and their response to the diet.

  5. When doing a double blind test with challenge materials, use several "provoking substances" together, not one at a time, as most have done.

  6. Perform an EEG both before and during diet therapy implementation, and to measure challenge results in the double-blind portion (Uhlig, Merkenschlager, Brandmaier, & Egger, 1997).

  7. Chronic candidiasis sensitivity has been implicated in ADHD (Kroker, 1987; Shaw, 1997). Pre-test the children for fungal metabolites and correlate levels with response to diet therapy.

  8. Studies on sulfation metabolism and the enzyme phenol sulfotransferase (PST) are opening a new arena for research into a possible cause of adverse reactions to food items that don't bother "normal" children (McFadden, 1996; O;Reilly & Waring, 1993; Sinaiko, 1996, Sinaiko, Lennett, Haus, & Edelkind, 1996; Waring, 1993). Pre-test the children with the Sinaiko MHPG urine test and correlate results with response to the diet. This is the next step in developing this test as a predictor of food-induced attention deficit disorder;


Much of the research is faulty, yet it seems that a substantial number of children do respond to dietary intervention, even allowing for exaggeration due to participant bias. If the 70% approximate improvement rate stands up to future research, then "food-induced ADHD" is not a subset of ADHD at all, but is the more common type of ADHD. If so, treatment paradigms need to be changed.


Adams, W. (1981). Lack of behavioral effects from Feingold diet violations. Perceptual and Motor Skills, 52, 307-313. MedLine

Armstrong, T. (1995). The myth of the ADD child , NY: Penguin Books.

Boris, M., & Mandel, F.S. (1994). Foods and additives are common causes of the attention deficit hyperactive disorder in children. Annals of Allergy, 72, 462-468. MedLine

Carter, C. M., Urbanowicz, M., Hemsley, R., Mantilla, L., Strobel, S., Graham, P. J., Taylor, E. (1993). Effects of a few food diet in attention deficit disorder. Archives of Disease in Childhood, 69, (5), 564-8. MedLine

FDA: 1993 DRAFT Redbook II: Toxicological principles for the safety assessment of direct food additives and color additives used in food.U.S. Food and Drug Administration.

Feingold Association of the United States. (1982; 1996; 1998) Handbook. Alexandria, VA: Author. More Info

Feingold, B. F. (1974). Why your child is hyperactive. NY: Random House, Inc.

Gross, M. D., Tofanelli, R. A., Butzirus, S. M., Snodgrass, E. W. (1987). The effect of diets rich in and free from additives on the behavior of children with hyperkinetic and learning disorders. Journal of American Academy of Child and Adolescent Psychiatry, 26, 53-55. MedLine

Harley, J. P., Ray, R. S., Tomasi, L., Eichman, P. L., Matthews, C. G., Chun, R., Cleeland, C. S., Traisman, E. (1978). Hyperkinesis and food additives: Testing the Feingold hypothesis, Pediatrics, 61, (6) 818-827. MedLine

Herbert, V. & Subak-Sharpe, G. J. (1995). Total nutrition: The only guide you'll ever need. Mount Sinai School of Medicine, NY: St. Martin's Griffin.

Hersey, J. (1996). Why can't my child behave? Alexandria, VA: Pear Tree Press. More Info

Holmes, D. S. (1997). Abnormal psychiatry. Addison-Wesley Educational Publishers Inc.

Kroker, G. F. (1987). Chronic candidiasis and allergy. In J. Brostoff, & S. J. Challacombe (Eds.), Food Allergy and Intolerance (pp. 850-872). London: Bailliere Tindall.

McFadden, S. A. (1996). Phenotypic variation in xenobiotic metabolism and adverse environmental response: Focus on sulfur-dependent detoxification pathways. Toxicology, 111, 43-65. MedLine

Medical Board of California v. Sinaiko, R. J. No. 13-93-28495 (filed Nov. 1, 1996; hearing Jan. 5 - Feb. 25, 1998).

NIH (1982). Defined diets and childhood hyperactivity. National Institutes of Health Consensus Development Conference Statement.

O'Reilly, B. A., & Waring, R. H. (1993). Enzyme and sulphur oxidation deficiencies in autistic children with known food/chemical intolerances. Journal of Orthomolecular Medicine, 8, 198-200.

PDR: Physicians desk reference (36th ed.). (1982) publ: Charles E. Baker, Jr. 1982.

Sinaiko, R.J. (1996, June). The biochemistry of attentional / behavioral problems. Paper presented at the annual conference of the Feingold Association of the United States, Orlando, FL. Full Text

Sinaiko, R. J., Lennett, P., Haus, B., & Edelkind, S. B. (1996). [Urinary excretion of MHPG conjugates in healthy children]. Unpublished raw data.

Swanson, J. M., Kinsbourne, M. (1980, March 28). Food dyes impair performance of hyperactive children on a laboratory learning test. Science, 207, 1485-1486. MedLine

Uhlig, T., Merkenschlager, A., Brandmaier, R., Egger, J. (1997). Topographic mapping of brain electrical activity in children with food-induced attention deficit hyperkinetic disorder. European Journal of Pediatrics, 156, (7), 557-561. MedLine

Waring, R. H., Ngong, J. M. (1993). Sulphate metabolism in allergy-induced autism: Relevance to the disease aetiology. Unpublished manuscript, Birmingham University, Edgbaston, Birmingham, UK.

Williams, J. I., Cram, D. M., Tausig, F. T., Webster, E. (1978). Relative effects of drugs and diet on hyperactive behaviors: An experimental study. Pediatrics, 61, (6) 811-817. MedLine