The development of brain imaging and machine learning technologies has allowed researchers to peer into the brain in real time, as well as discern hidden patterns in brain and behavioral data. For mental health professionals, these new technologies hold the promise that one day we will be able to ‘carve nature at the joints’: diagnose and evaluate the treatment of mental illnesses using objective neurological indicators of illness and improvement, rather than relying (as we do now) on observation and on clients’ subjective self reports.
This future vision however will not materialize easily, in part because research on the therapy-brain link is difficult to do well. For one, our current classification of psychological disorders (which is based mostly on reported symptoms) may not map well onto distinctly unique neurological patterns in the brain. Additionally, the link between brain neurology and mental illness is not simple. The brains of mentally ill people don’t usually reveal the corresponding, localized lesions seen in many neurological illnesses. Likewise, the inheritance patterns of mental illnesses are not straightforward. Most psychological disorders are caused by a combination of genetic predispositions, environmental conditions, and sociocultural factors, which is one reason for the dearth of useful animal models for most mental disorders.
Limitations notwithstanding, the new imaging and machine learning technologies have opened up intriguing lines of research into the complex ecology of mental health and illness. Much of this research has focused on three questions:
1: Does talk therapy cause changes in brain function and structure?
2: Can brain differences predict differences in client outcome?
3: Can differences in brains reliably diagnose specific psychological disorders?
Question 1: Does talk therapy cause changes in brain function and structure?
The notion that experience can affect the brain is not news. Some 70 years ago, Canadian researcher Donald Hebb proposed that brain neurology is plastic and responds to experience. As he famously put it: “The neurons that fire together wire together.” By the 1960s, research on rats had shown that enriched environments rewired the rats’ brains (see Rosenzweig & Bennett, 1996).
Accumulating data on humans have since converged to support this idea. For example, studying musicians, Gaser & Schlaug, (2003) found that, “gray matter volume differences in motor, auditory, and visual–spatial brain regions when comparing professional musicians…with a matched group of amateur musicians and non-musicians.” Studying London taxi drivers Maguire, Woollett & Spiers, (2006) found that, “compared with bus drivers, taxi drivers (who must pass a notoriously difficult street memorization exam in order to get licensed) had greater gray matter volume in mid-posterior hippocampi and less volume in anterior hippocampi. Furthermore, years of navigation experience correlated with hippocampal gray matter volume only in taxi drivers.” In a longitudinal study looking at the effects of mindfulness training on Parkinson’s Disease patients, Pickut et al., (2013) found a link between mindfulness based intervention and patients’ increased gray matter density in the neural networks postulated to play an important role in Parkinson’s Disease.
So, many kinds of experience have been shown to influence brain process and structure in many kinds of ways. But what about therapy experience, specifically? Suggestive evidence of the potential effects psychotherapy may have on clients’ brains has been accumulating for decades. For example, Baxter et al., (1992) measured resting-state brain metabolism in OCD patients before and after they received a course of either behavioral therapy or medication. The authors found decreased rates of glucose metabolism in the right caudate nucleus (an area involved in managing goal-directed behavior) of patients who responded to either type of treatment, compared with pre-treatment. Farrow et al., (2005) examined participants diagnosed with PTSD before and after CBT using an fMRI task that measured empathy and forgiveness. They found that, “the specific regions of the human brain activated by empathy and forgivability judgments changed with symptom resolution in PTSD.”
More recently, Månsson et al., (2016) found that left amygdala gray matter volume in socially anxious patients decreased significantly with Cognitive Behavioral Therapy (compared with a control treatment). The CBT-induced reduction of gray matter volume correlated positively with reduced anticipatory anxiety after treatment. Mason et al. (2016) found that symptom reduction after cognitive behavioral therapy for psychosis corresponded to changes in how the patients’ amygdala (a region linked to encoding emotional and threat related memory) were connecting with brain regions involved in self regulation.
Barsaglini et al., (2014) reviewed the pertinent literature on the effects of therapy on the brain across multiple disorders (obsessive-compulsive disorder, panic disorder, unipolar major depressive disorder, posttraumatic stress disorder, specific phobia, and schizophrenia). Longitudinal, peer reviewed studies were included if they used functional neuroimaging to examine the impact of psychotherapy on specific regions of the brain.
The authors noted that comparison among studies was made difficult by methodological inconsistencies, including the employment of different neuroimaging techniques, psychological interventions, and analytical approaches. Nevertheless, they concluded that, “based on the studies reviewed, it is clear that psychological interventions have the potential to modify brain function across a range of psychopathological conditions… Our review shows that a number of studies have indeed reported an association between changes in brain function and symptom improvement following the administration of psychotherapy… The results…support the notion that functional neuroimaging methods could potentially be used to evaluate the clinical impact of a particular treatment… However, this conclusion should be made with caution since only a small fraction of published studies (14%) included a waiting list, and were therefore able to dissociate between the direct impact of psychotherapy on brain function and the neurobiological changes associated with the natural course of the disorder.”
The authors also noted cogently that, “the primary aim of psychotherapy is to promote changes in the person’s mood, beliefs and behaviour, rather than the underlying functional changes in the brain… Such functional changes in the brain have in themselves no obvious meaning and need to be interpreted with reference to changes in the patient’s mood, beliefs and behavior… Thus, the use of neurobiological changes as an ‘objective’ means of monitoring the progress and outcome of treatment still relies on the ‘subjective’ reports by the patient and the therapist.”
Question 2: Can brain differences predict differences in client outcome?
It is a long time dream of psychotherapy researchers (and practitioners and clients) that a technology develops to allow us to chuck the ‘one size fits all’ and ‘trial and error’ approaches we currently use in applying psychotherapy in favor of a more precise and efficient system in which biological markers are used to tailor specific treatments to individual patients.
Research has pointed out the feasibility of this idea. For example Siegle et al., (2012) found that depressed participants with the lowest pre-treatment reactivity in the subgenual anterior cingulate cortex (a brain region involved in the modulation of emotional behavior) in response to negative words displayed the most improvement after cognitive therapy. Likewise Doehrmann et al., (2013) found that social anxiety patients whose brains reacted strongly to facial images (neutral vs. angry faces) before treatment benefitted more from the therapy. Specifically, reactions in two occipito-temporal brain regions (involved in processing of visual cues such as faces) correlated with positive cognitive behavioral therapy outcome.
Chakrabarty et al. (2016) reviewed 40 studies of predictive neuroimaging markers for treatment response in major depressive and anxiety disorders. The researchers noted that results across studies were highly variable. In part, this was due to the various limitations of the studies being reviewed. One common limitation was a small sample size. Small studies have little ‘statistical power,’ which means they may miss actual effects, in the same way that a weak flashlight may fail to illuminate a piece of furniture in the corner of a dark room. Small studies are also more susceptible to chance, in the same way that a bad free throw shooter is more likely to make 100% of his free throws if he gets to try three, as opposed to one hundred of them. Moreover, even studies that use the same methods tend to get different results, because participants who share a diagnosis may differ on many other, unmeasured variables that are nonetheless relevant to the effectiveness of treatment.
The researchers concluded that, “while the extant data suggest avenues of further investigation, we are still far from being able to use these markers clinically… Future studies need to focus on longitudinal testing of potential markers, determining their prescriptive value and examining how they might be integrated with clinical factors.”
Question 3: Can differences in brains reliably diagnose specific psychological disorders?
At the core, not much has changed in the diagnosis business in the realm of mental health since the days of Freud. Diagnosis still depends largely on signs (what we can observe in the client’s behavior) and symptoms (how the client says they feel). Signs and symptoms are important to be sure, but relying on them for diagnostic purposes is problematic. For one, by the time mental illness manifests itself in visible signs, it may be late in the game to intervene effectively. Moreover, people are often unreliable reporters of their own experience.
Again here, many researchers and psychologists dream of a day when some brain imaging technique may be able to diagnose mental illness with the same level of reliability and ease that a simple blood test can today diagnose, say, HIV. In light of advances in machine learning and computing, many hope that a computer, fed with brain imaging data from thousands of patients, may soon be able to find the patterns that reliably and objectively separate, for example, the brains of depressed patients from those of socially anxious ones.
Research in this area has shown promise. For example, Bansal et al., (2012) used MRI datasets from patients with various disorders (Attention-Deficit/Hyperactivity Disorder, Schizophrenia, Bipolar Disorder, etc.) to train a semi-supervised learning algorithm to discover natural groupings of brains based on “the spatial patterns of variation in the morphology of the cerebral cortex and other brain regions.” The method discriminated with high precision the brains of persons who had one specific neuropsychiatric disorder from the brains of healthy participants and the brains of persons who had a different neuropsychiatric disorder. Bedi et al., (2015) used automated speech analyses combined with machine learning to derive speech features (including a measure of semantic coherence and two syntactic markers of speech complexity) that predicted patients’ later psychotic episode with 100% accuracy, outperforming predictions from clinical interviews.
Results of this kind are promising. However, reviewing the research literature, Savitz et al., (2013) noted the difficulties. First, psychological disorders are highly heterogeneous; there is considerable overlap between patients and healthy controls in neuroimaging measures. For example, healthy individuals with a family-history of mood disorders show similar neuroimaging abnormalities as ill patients, even if they are not depressed themselves (and may never become so). Additionally, neuroimaging techniques—especially fMRI—are highly sensitive to normal, time-related fluctuations in patient physiology. Medical conditions that commonly co-occur with mood disorders may also affect imaging data. Medications (or other substances) are potent confounds because they may affect brain structure and function, thus biasing classification algorithms. Noting these limitations, the researchers concluded that, “Although there are a number of promising results… there are currently no brain imaging biomarkers that are clinically useful for establishing diagnosis…in mood disorders.”
In summary, as it stands, the answers to the three questions posed at the beginning of this article seem to be partial and conditional at best: Does talk therapy cause changes in brain function and structure? Probably yes, but we’re not sure how, and what it means. Can brain differences usefully predict differences in client outcome? Not quite yet, but maybe in the future, at least for some disorders. Can brain differences usefully diagnose specific psychological disorders? In theory (and simulation), yes, but in the clinic and hospital not quite yet, and probably not for a while.