Reader Comments

Post a new comment on this article

Temperature measurements

Posted by bryanpaton on 03 Nov 2011 at 23:22 GMT

A very interesting article but two issues that I and fellow colleagues have encountered during temperature measurements during the RHI:

1. The Fluke series of infra-red thermometers, one particular model of which you used, have a stated accuracy of +/- 2 degrees F (1.5 degrees for the highest spec. model) and a repeatability (precision) of +/- 2 degrees F.

http://www.fluke.com/fluk...

Averaging will help with precision but it cannot change accuracy (only calibration will do that), also using difference scores will alleviate issues of accuracy. Given all of this and that these specifications are given under ideal conditions how reliable can measurements be if they are reported an order of magnitude less than the specifications?

2. The video linked in the article indicates how the experiment was performed. In particular it shows the distance at which the infra-red thermometer was used. Most such thermometers incorporate a lens and focusing apparatus. In addition the sensors themselves have a particular field of view (FOV). Given that the area the FOV encompasses increases with distance and that all such infra-red sensors report the average temperature across the FOV when taking measurements you are not only measuring the object or spot centred by the targeting laser (or similar) but also everything else inside the cone (defined by the FOV and the distance to the object). This will often include the table top or experimental apparatus and if held far enough away could include the other arm.

No competing interests declared.

RE: Temperature measurements

bryanpaton replied to bryanpaton on 03 Nov 2011 at 23:24 GMT

I neglected to add an actual question for point 2. Given the above how sure can you be that the measurements you took accurately reflect the temperature of the arm/hand exclusively?

No competing interests declared.

RE: RE: Temperature measurements

parks1 replied to bryanpaton on 04 Nov 2011 at 00:21 GMT

Thank you very much for your comments.

First of all, the video on our website is definitely NOT the step-by-step method of how it was done. Rather, it is a just a cursory illustration. In the actual experiment, we took great care to standardize the measurement distance and angle. The procedure always required two experimenters.

However, you are correct in pointing out the limits of the thermometer. On the other hand (not intended to be a pun!), the temperature changes are consistent across participants and affects the two hands differentially.

Moreover, we conducted a separate control experiment ( see Supplemental data) to measure the temperature of the surfaces on which the subjects' hands were resting both inside and outside of the box.

There are further data on the temperature measurements under "Supplementary Data" section at the end of this article. It is also copied below.

S1. Supplementary Data

To examine the possibility that skin temperature changes were due to specific conditions of our experimental set-up (e.g. different surfaces upon which subjects placed their stimulated and unstimulated hands) rather than to RHI induction, we performed a control experiment on a sample of 20 healthy undergraduate and graduate students (9 M / 11 F; age: 24.1 ± 4.4). The experimental apparatus and procedure was identical to the main study. However, no rubber hand was placed in the compartment adjacent to the real hand, and no stimulation was applied to the real hand. Temperature was measured at the same three points on both the hand inside the box, which would have been the stimulated hand, and the hand outside of the box, which would have been the unstimulated hand. These temperature measurements were taken at ‘baseline’ and 1,2, and 3 minutes following baseline. This procedure was repeated with the opposite hand inside the box; hand order was counterbalanced across subjects. Again, temperature was averaged across the three locations on both the hand inside and hand outside the box. The change in temperature for each hand was quantified as the difference in temperature between baseline and the average of the three post-baseline time points. A negative change indicated cooling relative to baseline.
A repeated measures ANOVA was performed on temperature change with position (inside or outside box) and target hand (hand placed inside the box) entered as factors. No main effect of target hand (F (1,19)=0.95, p=0.34) or target hand-by-position interaction effect (F (1,19)=1.45, p=0.24) was observed. There was a trend towards a main effect of position (F (1,19)=3.74, p=0.07), with a larger increase in temperature from baseline in the hand that was placed in the box (inside box: 0.52 ± 1.36°; outside box: 0.21 ± 1.33°). In contrast, during the main experiment the temperature of the stimulated hand, which was placed inside the box, decreased significantly; the results of this control study suggest that this pattern of cooling of the stimulated hand and warming of the unstimulated hand cannot be accounted for by the nature of the surface on which the two hands were placed.
Baseline temperatures in this control experiment were also examined. A repeated-measures ANOVA was performed on baseline temperature, with hand and position entered as factors. Replicating the finding from the main study, there was a significant main effect of hand (F (1,19)=12.52, p=0.002), with the left hand being warmer than the right hand. However, there was no significant main effect of position (F (1,19)=0.22, p=0.64) or hand-by-position interaction (F (1,19)=0.005, p=0.95). In the main study, the temperature of the stimulated hand, which was positioned inside the box, was warmer than the unstimulated hand at baseline. The lack of a significant position effect in this control study indicates that greater temperature in the stimulated hand at baseline might have been due to task-specific, rather than environment-specific, factors.
To summarize, the results of this control experiment indicate that the limb-specific temperature modulations observed in the main experiment were not caused by idiosyncratic parameters of the experimental environment, but instead, were related to RHI induction.

No competing interests declared.

RE: RE: RE: Temperature measurements

bryanpaton replied to parks1 on 04 Nov 2011 at 01:11 GMT

Thank you for the reply. I did read the supplementary material before posting my query. Of note the data in the supplementary material and the article itself does not address the second issue I raised. Standardising the distance and angle is good and expected but unless the thermometer was held close or against the skin then it was not just measuring the point of interest but also the surrounds. Even averaging across locations will not eliminate this, especially as the mean is very sensitive to outliers.

In regards to my first point it is entirely possible to obtain consistent outcomes even if you are measuring noise.

No competing interests declared.

RE: RE: RE: RE: Temperature measurements

parks1 replied to bryanpaton on 04 Nov 2011 at 19:57 GMT

The Fluke thermometer, when it is switched on emits a low-level laser beam so that the experimenter can aim this laser beam at the center of the area of measurement. According to the manufacturer (Fluke Corporation, WA), the “spot” size that indicates 90% encircled energy is about 3.8cm at a distance of about 30cm, and that the target area should be larger than the spot size so that the temperature of the area being measured falls within that limit. The “spot” size of the area being measured by the thermometer becomes larger as the distance between the thermometer and the target surface increases (similar to the calculation for the visual angle).

The average length of an adult male hand is reported to be about 19cm, while the average length of an adult female hand is about 17cm. The average hand width for adult males and females is 8cm and 7cm, respectively. (Agnihotri, Purwar, Jeebun, & Agnihotri, 2006). We marked three areas on the back of the hand with a washable marker so that we could aim the laser at the marked dots. Thus, we sampled three areas from each hand (below the second and fifth digits and on the wrist), marked with washable marker and averaged these temperature measurements every time.

The “spot” size that indicates 90% encircled energy is about 3.8cm at a distance of about 30cm, and an average male hand is about 19cm-by-8cm. The distance between the hand and the thermometer was kept at about 30cm (length of a school ruler). These parameters suggest that the areas that we measured fall within the “spot” size recommended by the manufacturer.

Although we tried to keep the distance between the thermometer and the hand the same for all measurements across all subjects, we did not have a fixed stand for the thermometer so sometimes the thermometer might have been closer than 30cm and sometimes a little farther than 30cm. But for us, it was extremely important to get the measurements done before the RHI effect wore off, so we aimed to keep the distance roughly 30cm and take the temperature as soon as possible.

Future studies could employ a more standardized procedure for temperature measurements or use more sophisticated equipments. However, it may not be ideal to attach sensors to the hand to improve temperature measurements because they could interfere with the tactile experience.

Moreover, the major goal of this study was to examine body ownership using multiple behavioral measures, with the temperature change being one of them. The drift and the questionnaire measures were very easy to collect and yielded robust results. So perhaps, the temperature measures are not essential for the purpose of studying altered body ownership and agency in this particular population.

Reference
A. K. Agnihotri, B. Purwar, N. Jeebun, and S. Agnihotri. Determination of sex by hand dimensions. The Internet Journal of Forensic Science, 1(2), 2006.

No competing interests declared.

RE: RE: RE: RE: RE: Temperature measurements

bryanpaton replied to parks1 on 10 Nov 2011 at 05:28 GMT

We came to similar conclusions regarding skin temperature and the RHI as well as the use of thermistors placed directly on the skin. The thermistors can be orders of magnitude more precise and accurate but the benefits might not outweigh the potential confounds.

Sorry to belabour the point but in regards to the precision (which is important for repeated measures measurements) the state precision is an order of magnitude larger than the temperature differences you are reporting. Whilst as I mentioned precision (think of it as the Standard Deviation) can always be increased (SD decreased) by collecting more samples. However as you would remember from the central limit theorem the sample SD is related to the population SD (in this the desired precision level) by:

desiredSD = sample SD / Square root (N samples)

which means to get an actual precision of 0.1 F (ie to reliably talk about or express effects to that level of precision) each "trial sample" ie the number for each participant for a given trial it would have to be the result of the averaging of 100 samples (square root of 100 is 10, sample SD is always 1 F).

No competing interests declared.