The science of solitary confinement

CBC’s Quirks and Quarks had a seg­ment this week about the psy­chol­o­gy of soli­tary confinement.“Months locked in a tiny box—how soli­tary confinement can erode men­tal health”, Quirks and Quarks, CBC (8 February 2019). Bob McDonald in­ter­viewed Dr. Craig Haney, one of the ex­pert wit­ness­es in the BC Supreme Court case that re­sult­ed in por­tions of Canada’s soli­tary-confinement regime be­ing struck down as un­con­sti­tu­tion­al.British Columbia Civil Liberties Association v Canada (Attorney General), 2018 BCSC 62 [BCCLA v Canada].,Canada appealed that result. The appeal was heard on November 13, 2018. The decision had not been announced at the time of this blog post’s original publication, but on June 24, 2019, the Court of Appeal for British Columbia held that “a legislative provision that authorizes the prolonged and indefinite use of administrative segregation in circumstances that constitute [solitary confinement]” are an unjustified violation of s 7 of the Charter. The Court of Appeal accepted all of the trial judge’s factual findings, which are the subject of this blog post. British Columbia Civil Liberties Association v Canada (Attorney General), 2019 BCCA 228.

This is a short com­ment on that case, British Columbia Civil Liberties Association v Canada (Attorney General) (BCCLA v Canada), fo­cus­ing on the judge’s han­dling of the scientific ev­i­dence.

Context

This case turned on the judge’s un­der­stand­ing of nu­anced scientific ar­gu­ments be­tween ex­perts with com­pet­ing opin­ions. The scientific ev­i­dence was es­sen­tial to the judge’s finding that the way that we cur­rent­ly use soli­tary confinement “places all Canadian fed­er­al in­mates sub­ject to it at significant risk of se­ri­ous psy­cho­log­i­cal harm, in­clud­ing men­tal pain and suffering, and in­creased in­ci­dence of self-harm and sui­cide.”BCCLA v Canada, supra note 3 at para 247.

I did not lis­ten to or read tran­scripts of the tes­ti­mo­ny of the scientific ex­perts in this case. This com­ment is based only on the pre­sen­ta­tion of the ev­i­dence by Justice Leask in his writ­ten de­ci­sion. In the fol­low­ing, I as­sume that he has giv­en a fair pre­sen­ta­tion of that ev­i­dence. I am not a psy­chol­o­gist, but I am a sci­en­tist. I un­der­stand ev­i­dence, ex­per­i­ments, sta­tis­ti­cal tests and pow­er, cau­sa­tion, and the strengths and weak­ness­es of meta-analy­ses and sys­tem­at­ic re­views.

My take on the evidence as presented

Plaintiff’s experts

The ev­i­dence sup­ports a claim that soli­tary confinement caus­es var­i­ous harm­ful effects.Ibid at paras 163–194. There is a large body of clin­i­cal and ex­per­i­men­tal re­search es­tab­lish­ing that when you re­strict a per­son’s en­vi­ron­men­tal and so­cial stim­u­la­tion, they suffer. Dr. Grassian has ob­served, in pris­on­ers in soli­tary confinement, a rare confluence of psy­chi­atric symp­toms which we al­ready know can be caused by re­strict­ing en­vi­ron­men­tal and so­cial stim­u­la­tion. There is a cor­re­la­tion with the on­set of the in­ter­ven­tion: in­tro­duc­tion of soli­tary confinement pre­dictably shifts EEG pat­terns to­wards that char­ac­ter­is­tic of stu­por and delir­i­um. Psychiatric dis­tur­bances spring up anew when pris­on­ers are seg­re­gat­ed. When the seg­re­gat­ed confinement is pro­longed, there is a high­er in­ci­dence of psy­chosis.

Further (and this re­lates to how we should view the gov­ern­ment’s ev­i­dence), the na­ture of soli­tary confinement and the psy­chi­atric symp­toms make self-re­port­ing a poor mea­sure of effects. Prisoners will avoid ac­knowl­edg­ing psy­cho­log­i­cal harm out of spite against au­thor­i­ties whom they see as try­ing to “break them down”. Psychotic dis­tur­bances of­ten have a dis­so­cia­tive char­ac­ter: in­di­vid­u­als of­ten do not re­call events which oc­curred dur­ing con­fu­sion­al psy­chosis.

Dr. Grassian’s ev­i­dence must be seen in the con­text of the Bradford Hill cri­te­ria (use­ful for as­sess­ing cau­sa­tion from epi­demi­o­log­i­cal ev­i­dence). In the case of soli­tary confinement, it would not be eth­i­cal to per­form a ran­dom­ized con­trolled tri­al. The gov­ern­ment’s ex­perts, Dr. Mills and Dr. Gendreau, point to the in­abil­i­ty to de­duce cau­sa­tion from this re­search. While this may be true, due to the lack of di­rect ex­per­i­men­tal ev­i­dence, the Bradford Hill cri­te­ria give guid­ance for mak­ing (ab­duc­tive) in­fer­ences in spite of this.

The harm­ful effects are con­sis­tent across pop­u­la­tions, the effects hap­pen af­ter the pur­port­ed cause, there is a bi­o­log­i­cal gra­di­ent (longer ex­po­sure is as­so­ci­at­ed with high­er in­ci­dence), there is a plau­si­ble mech­a­nism for the cau­sa­tion, and the hy­poth­e­sis is con­sis­tent with lab­o­ra­to­ry ex­per­i­ments on de­pri­va­tion of en­vi­ron­men­tal and so­cial stim­u­lus. This confluence of ev­i­dence sup­ports an ab­duc­tive in­fer­ence that soli­tary confinement caus­es the harm­ful effects.

Government’s experts

The con­clu­sions drawn by the gov­ern­ment ex­perts are not con­sis­tent with the ev­i­dence they claim to base it on. They look to low-pow­er ob­ser­va­tion­al stud­ies that used sub­jec­tive self-re­port­ing as a mea­sure. These stud­ies have a poor abil­i­ty to de­tect an effect even if an effect were to be present. Because these stud­ies failed to find an effect, the gov­ern­ment ex­perts con­clud­ed that there was no effect (“seg­re­gat­ed in­mates with­out men­tal ill­ness do not ex­pe­ri­ence de­bil­i­tat­ing psy­cho­log­i­cal or psy­chi­atric symp­toms due to their place­ment in seg­re­ga­tion”), or an effect of low­er mag­ni­tude than ar­gued by the plaintiffs (“that it is much milder than that pre­dict­ed by the plaintiffs’ ex­perts”), but these con­clu­sions can­not be drawn from a neg­a­tive re­sults in low-pow­ered stud­ies.

Suedfeld et al (1982), a study re­lied upon by the gov­ern­ment, gives con­tra­dic­to­ry and weak ev­i­dence.Ibid at para 200. “A re­la­tion­ship was found be­tween the length of stay and mea­sures of de­pres­sion and hos­til­i­ty” (a dose-re­sponse re­la­tion­ship; one of the Bradford Hill cri­te­ria that points in favour of finding an effect). But be­cause of the small mag­ni­tude of the ob­served effect, the au­thors of that study gave a mea­sured con­clu­sion: it “did not sup­port the claim that soli­tary confinement […] was over­whelm­ing­ly aver­sive, stress­ful, or dam­ag­ing to in­mates.” Again, this is just fail­ure to find ev­i­dence for the effect, not pos­i­tive ev­i­dence against the effect.

The gov­ern­ment ex­perts pushed the court to rely on lon­gi­tu­di­nal stud­ies, which are more like an ex­per­i­ment in that they ob­serve what hap­pens be­fore and af­ter the on­set of an in­ter­ven­tion. They be­lieve the “Zinger study” to be worth significant weight.Ibid at paras 203–204. The pop­u­la­tion for this study was 83 seg­re­gat­ed plus 53 gen­er­al-pop­u­la­tion pris­on­ers. The seg­re­gat­ed pop­u­la­tion ini­tial­ly re­port­ed more de­pres­sive symp­toms, but not on a mea­sure of hope­less­ness. Both seg­re­gat­ed and non-seg­re­gat­ed pris­on­ers im­proved over the 60 days of the study on mea­sures of de­pres­sion, anx­i­ety, hope­less­ness, and psy­cho­log­i­cal ad­just­ment. The fact that both pop­u­la­tions im­proved over the 60-day study is in­ter­est­ing scientifically. It puts into ques­tion whether the fact of ob­ser­va­tion on its own had an effect on the pop­u­la­tion or on their re­port­ing. Perhaps when pris­on­ers re­ceive in­creased at­ten­tion from oth­ers re­gard­ing their well-be­ing and are asked to per­form mean­ing­ful in­tro­spec­tion and tell some­one about, they feel bet­ter. Or maybe the par­tic­i­pants felt a sub­con­scious pres­sure to re­port im­prove­ment. Without blind­ing, this kind of ob­serv­er effect is com­mon.

Of course, a lon­gi­tu­di­nal study that tracks the re­sponse to an in­ter­ven­tion over time has the po­ten­tial to be more use­ful scientifically than a set of clin­i­cal ob­ser­va­tions as pre­sent­ed by Dr. Grassian, but the Zinger study has method­olog­i­cal weak­ness­es that make its neg­a­tive finding un­sur­pris­ing and definitely not pos­i­tive ev­i­dence for the claim that there is no effect.

Another study that the gov­ern­ment ex­perts em­pha­sized was a pair of meta-analy­ses by Morgan and Gendreau. Meta-analy­ses (and sys­tem­at­ic re­views es­pe­cial­ly) can be a great tool for eval­u­at­ing an en­tire body of ev­i­dence gen­er­at­ed by many sep­a­rate re­search teams. By pool­ing to­geth­er par­tic­i­pants, meta-analy­ses can find effects when in­di­vid­ual stud­ies were too small to find one. They can tease out pat­terns of a de­creas­ing effect size with in­creased rig­or (a mark­er of there be­ing only a small or no effect). They can de­tect pub­li­ca­tion bias (where re­search teams across an en­tire field with­hold stud­ies that don’t show sta­tis­ti­cal­ly significant effect).

Paragraph 206 sum­ma­rizes the Morgan and Gendreau stud­ies: “the two stud­ies […] found that the differences be­tween seg­re­gat­ed and non-seg­re­gat­ed in­mates ranged from small to mod­er­ate on a wide range of psy­cho­log­i­cal in­di­ca­tors.”Isn’t this actually a finding in favour of an effect? But, they also found that “the size of these differences were significantly re­duced when only the stud­ies rat­ed as more scientifically rig­or­ous were con­sid­ered.” As I men­tioned above, this pat­tern in a body of re­search is an in­di­ca­tor of a small or non-ex­is­tent effect. But, us­ing this move de­pends on an ac­cu­rate as­sess­ment of a study’s scientific rig­or, and Dr. Grassian, one of the plaintiff’s ex­perts, crit­i­cized the se­lec­tion cri­te­ria for the Morgan study.BCCLA v Canada, supra note 3 at para 234.

Regardless, the gov­ern­ment ex­pert con­cludes: “the findings do not sup­port the no­tion that, as a group, in­mates in seg­re­ga­tion suffer […] when com­pared with in­mates out­side of seg­re­ga­tion.”

Two things are im­por­tant from this con­clu­sion. It is a con­clu­sion “as a group”.Ibid at para 207. And again, even tak­en at face val­ue, it is only a fail­ure to find pos­i­tive ev­i­dence for the effect, not pos­i­tive ev­i­dence that there is no effect. With an effect that might vary with­in spe­cial­ized sub­pop­u­la­tions, and sub­jec­tive psy­cho­log­i­cal mea­sures in­volv­ing self-re­port­ing that the par­tic­i­pants may have mo­ti­va­tions or in­cen­tives for hid­ing, it isn’t un­ex­pect­ed that a meta-analy­sis would find no ev­i­dence for an effect.

Last, the gov­ern­ment re­lies on the “Colorado Study”. But, there is one crit­i­cism of the study that is so fa­tal to its rel­e­vance that I will not even present its re­sults. The study re­lied on self-re­port­ing scales. But, the cor­rec­tions staff also kept ob­jec­tive data in the form of psy­chi­atric cri­sis records (for self-harm, sui­cide, psy­chot­ic dis­tur­bances). These cri­sis records square­ly con­tra­dict­ed the pris­on­ers’ self-re­port­ing. According to the self-re­port scales, there was no de­te­ri­o­ra­tion with­in the seg­re­gat­ed pop­u­la­tion over the course of 12 months. The cri­sis records show the op­po­site. Only 12 episodes oc­curred in the first six months; 25 oc­curred in the sec­ond six months.Ibid at para 237.

The method­olog­i­cal weak­ness­es in both the “Colorado study” and the “Zinger study” also affect the rel­e­vance of the Morgan and Gendreau meta-analy­sis be­cause they were both in­clud­ed in the small set of stud­ies meet­ing the se­lec­tion cri­te­riaIbid at para 234., and the “Colorado study” was one of the stud­ies identified to be weight­ed as hav­ing a “stronger” qual­i­ty de­sign.Ibid at para 243.

Discussion

This is a difficult scientific ques­tion. The plaintiff’s ex­perts do not point to ev­i­dence that ex­per­i­men­tal­ly es­tab­lish­es a causal link be­tween soli­tary confinement and psy­cho­log­i­cal harm. They do, how­ev­er, point to a confluence of ev­i­dence that, when viewed in light of the Bradford Hill cri­te­ria, strong­ly sug­gests an effect.

The ev­i­dence pre­sent­ed by the gov­ern­ment is based on meth­ods that have the po­ten­tial to be more re­veal­ing. However, due to method­olog­i­cal weak­ness­es in­volved in the specific ap­pli­ca­tions of those meth­ods, this ev­i­dence is ac­tu­al­ly not re­veal­ing.

I have em­pha­sized the dis­tinc­tion be­tween (1) fail­ing to find an effect and (2) finding pos­i­tive ev­i­dence that there is no effect. Of course, one can­not prove a neg­a­tive; fail­ure to find pos­i­tive ev­i­dence that there is no effect should not be fa­tal to any counter-ev­i­dence. But, when you present a study that you in­tend to negate the claim that there is an effect, that study should at least have been set up to be able to find an effect if the effect ex­ists. That is, the study should be high pow­ered. Failure to find an effect in a high pow­er study can be very com­pelling ev­i­dence.

That is a tricky dis­tinc­tion and I am im­pressed that Justice Leask saw it. In this case, ev­i­dence that was ob­tained by meth­ods that are gen­er­al­ly less ro­bust was owed more weight than ev­i­dence ob­tained by meth­ods that are gen­er­al­ly more ro­bust.Ibid at para 254.

In a Section 7 chal­lengeThat is, when a plaintiff challenges a government law or act on the basis that it infringes upon their Section 7 right to life, liberty, or security of the person., the plaintiffs only need to show a “sufficient causal con­nec­tion” be­tween the gov­ern­ment ac­tion and the harm. They don’t need to show that soli­tary confinement is “the only or the dom­i­nant cause” of the harms. And, they only need to es­tab­lish “a rea­son­able in­fer­ence, drawn on a bal­ance of prob­a­bil­i­ties.”Canada (Attorney General) v Bedford, 2013 SCC 72 at para 76. Justice Leask found that the plaintiffs met that bur­den.

Related media

“End Solitary Confinement–BobbyLee’s Story”, BC Civil Liberties Association (26 May 2016).

Susan Haack, Defending Science–with­in Reason (Amherst, New York: Prometheus Books, 2007).

Susan Haack, “Of Truth, in Science and in Law” (2008) 73:3 Brooklyn Law Review 985.

Craig Haney, “Restricting the Use of Solitary Confinement” (2017) 1 Annual Review of Criminology 285.

Treacy Ziegler, “Memory of Space”, The Prison Arts Coalition (15 October 2014).

Treacy Ziegler, “The art of ab­solute lone­li­ness”, Broad Street Review (3 June 2014).