Experiments | Contrast
The contrast is an important part of the experimental design. It describes the key differences that are being compared within an experiment. A general example would be "Task > Rest", where the contrast is the difference between the active task and rest. This section describes the specific changes that were analyzed to create an experiment's SPI.
Possible aspects of a contrast include:
- Stimulus Modality
- Stimulus Type
- Response Modality
- Response Type
- External Variable
- Subject Group
- Experience-Dependent Change
Experiments can have more than one aspect being contrasted. Around half of the experiments in the functional database have only one thing changed, a quarter have two, a quarter have three or more. Rarely (~1%), an experiment will have six or more contrasted features. The types of contrast involved depends heavily on the details of the experimental design. Find more information in an example experiment with four aspects being contrasted.
A contrast of stimulus modality means that the broad modality of the stimulus has changed between conditions, and those conditions are being compared in this experiment. For example, face discrimination versus rest, which changes both stimulus modality and type. Often when the stimulus modality changes, the stimulus type does too, but this is not always the case. One example of that is visually presenting a written stimulus versus hearing the same stimulus read aloud.
A change in stimulus type could be an experiment that compares viewing letters versus numbers, or hearing tones versus music. Experiments which compare stimulus to no stimulus count as changing stimulus type.
Stimulus type is by far the most common change in a contrast. Around 70% of the database includes a contrast of stimulus type, followed by instructions around 45%, with all others around 20% or below.
An experiment which compares a condition with a response to one with no overt response, like fixation or rest, includes response modality and type in its contrast. Occasionally a contrast will involve the modality but not the type, like a button press or a flexion/extension that is completed with different body parts.
A contrast that involves a difference in response modality would be the following: a study on children with fetal alcohol syndrome (FAS) that looks for differences in activation pattern when the method of response changes. Specifically, it compares functional activity patterns in the cerebellum associated with eye-blinking vs. finger-tapping. As such, a classic eye-blink conditioning paradigm was set-up and as expected the majority of the FAS children showed the conditioned eye-blink response just prior to the air puff. In a follow-up study these same FAS patients were invited back and using a metronome at first and then on their own, were asked to perform a rhythmic finger-tapping task. Cerebellar activation patterns were compared across eye-blinking and finger-tapping.
A contrast that involves different response types could be as follows, a study explored differences in how our brain reacts when we respond to stimuli via button-press or by pointing. Thus, subjects were asked to indicate which of the two words presented on a screen, were included in the list of words they saw previously. Depending on the trial, some subjects indicate their choice by pressing the left key for the leftmost word and the right, for the rightmost. Other subjects are asked to indicate choice by simply pointing at the appropriate word. Notice that the response modality of “Hand” is unchanged, since subjects use their hands to both depress certain buttons or point to indicate choice.
A contrast of instructions means that the compared conditions had a difference in the way the subjects were told to perform.
For example in Anderson, et al. JCN 2000 (BMapID 30001), there were four conditions in which subjects were presented with visual words and auditory tones and responded with a spoken word. The first condition showed subjects paired words and instructed them to encode the words, then read the second word aloud. The second task showed the first of a pair of words along with "word" and subjects were instructed to speak the missing word or say "pass". The third and fourth conditions investigated divided attention by continuing the paired word task as before, but adding in the additional instructions of pressing one button when a low tone was heard and another button when a high tone was heard. This experimental design has avoided a change in stimulus modality and stimulus type by presenting the words and tones throughout all the conditions. The instructions, however, change between "encode and read words", "recall and speak words" and "encode and recall words while also distinguishing tones".
An external variable can be used to inform a regression analysis which finds the voxels in which neural activity was most correlated (either positively or negatively) with the measured variable.
For example, "Positive Correlation with Syllable Rate" would create a statistical parametric image by finding the degree of positive correlation between each voxel's activity and the measured syllable rate, across all subjects' scans and responses. This can be co-coded with another contrast if it applies. So if more than one subject group used to make the correlation image, then the contrast would also include Subject Group.
Group contrasts are where the difference is not in the conditions experienced, but in the subject groups which undergo it. A general example would be "Patients > Normals", and this is most commonly how it is used, but group contrasts could also study gender ("Females > Males"), aging ("Children - Adults"), handedness ("Right > Left"), native language ("Chinese - English") or anything else that can be explored by comparing different populations.
When the conditions being compared were acquired in different time periods, this is contrasting based on session. Often multi-sessions experiments are longitudinal studies. These could be investigating the progression of a disease ("Baseline > Followup in Alzheimer's patients"), rate of recovery ("14 Months after Stroke > 2 Months") or treatment effects ("Post-Therapy > Pre-Therapy").
Session contrasts could also study learning ("Post-Conditioning > Pre-Conditioning"), drug effects ("Pre > Post Quit in Smokers") and more ("Pre > Post Sleep Deprivation").
Experience-dependent change is the least commonly used contrast in the database with only 2% of papers, followed by session-based contrasts at 4%. These two contrasts are very similar; while sessions are comparing before and after events that happen outside of the study, experience-dependent change compares before and after events that happen during the study itself. This most commonly occurs when subjects perform the same conditions repeatedly and are expected to improve with practice that occurred during the earlier tasks ("Third Run > First Run"). Sometimes the experience is not purely practicing a task, but subjects learning about the experimental design in a way that effects performance ("with decreasing reward").
The description field is used to summarize and clarify the contrast of a given experiment. For example, if the contrast includes External Variable, it should include the name of the relevant external variable. When multiple contrasts are selected, provide a brief summary of the experimental design.
To illustrate, let’s look at an example with multiple contrasts. Frodl (et al., 2007) examined MDD patients on either the antidepressant venlafaxine or mirtazapine. Subjects were scanned prior to treatment (baseline) and a follow-up four weeks later. Stimuli were randomized: presented stimuli were triplets of male/female emotional faces (8 blocks) or control stimuli were geometrical figures (9 blocks); faces were presented with one face at the top of the screen and two, at the bottom. Using a two-button press, subjects were instructed to choose which of the faces at the bottom matched the gender of the face at the top (the left or right face). The following “Experiment” involved an interaction analysis wherein severity of depression, as measured by the Hamilton Depression Rating Scale (HDRS), was added as a covariate in the computed ANCOVA. The results of which found significant increases in activation, from baseline to follow-up for the mirtazapine group. Thus, the coder selected 4 different contrasts:
- Stimulus Type: Male/Female faces vs. Geometrical figures
- Instruction: “Choose which face at the bottom matches the same gender of the face at the top” vs. “Determine which figure at the bottom matched the shape of the target figure on top”
- Session: 4 weeks (post-treatment) vs. Baseline
- External Variable: HDRS scores before and after treatment
Notice here that the Stimulus Modality was not selected because the way in which stimuli were presented did not change. Also, neither the response type or response modality was selected since the two-button press was also used for the control task. Similarly, Experience-Dependent Change would not apply because the subject’s experience, or rather the requirements of the task did not differ from group to group. Lastly, the Group itself would also not apply here because all subjects in this contrast were from the same medication group.