GBA Logo horizontal Facebook LinkedIn Email Pinterest Twitter Instagram YouTube Icon Navigation Search Icon Main Search Icon Video Play Icon Plus Icon Minus Icon Picture icon Hamburger Icon Close Icon Sorted
Guest Blogs

What is Thermal Bridging?

Why R-19 is never R-19

The studs and diagonal bracing are darker than the insulation between the studs because they are colder. Framing lumber is a better conductor than fiberglass insulation, so it bleeds heat through the wall assembly.

Like wind washing, thermal bridging is something folks mention all the time during audits (meaning they never ask about it). But what is thermal bridging, and why do I keep bringing it up when my customers just want new windows?

To understand thermal bridging, you need to understand your home’s wall assembly and the various materials used in its construction.

My house is well insulated, right?

The exterior walls of your home are (hopefully) well insulated. The building shell of your home provides the structural barrier between your dream home’s interior and the rest of the world. It keeps out unwanted air infiltration (mostly), moisture (again, fingers crossed) and retards the flow of heat through the wall. If well insulated, your home ought to stay warmer during cold winter months and cooler in the summer.

R-value is the measure of how well your insulation resists heat flow across the material in question. (For more information on R-value, see Understanding R-Value.) In theory (quite literally — a material’s theoretic R-value is tested under lab conditions), R-value is supposed to be an objective comparison of the thermal resistance of different materials.

Unfortunately, this just isn’t so. As I explained in a previous post, we all wish there was a magic, universally applicable number like car gas mileage with which to compare various insulation. But there isn’t.

Thermal bridging is one reason you can’t blindly judge insulation based on its R-value.

Standard wall assembly construction (at least in the Northeast) consists of drywall on 2×6 studs, 6-inch-thick fiberglass batts within the wall cavities, plywood or OSB sheathing, Tyvek building wrap as a water control plane, and siding. The R-value for any cross-section of the wall can be found by adding the material R-value of each layer.

A boring definition of thermal bridging

Thermal bridging occurs when a more conductive (or poorly insulating) material allows an easy pathway for heat flow across a thermal barrier. The most common form is probably within the eyesight of every reader of this article: wall studs.

Suppose your walls have 6″ fiberglass batts. (I’m sorry. We’re working to update the building code.) Every 16 inches on center in that wall is a 2×6 or 2×4 stud. The fiberglass has an R-value of around 3.5 per inch, and the stud is around R-1.2 per inch. The wood studs allow heat to flow through the wall assembly at a rate that is 3 times faster than the heat flow through the surrounding insulation.

While the advertised R-value for a 6-inch fiberglass batt is R-19, the building assembly’s effective R-value is about R-3 lower. Bummer.

Cures for the thermal bridging blues

The question then becomes… What to do?

Simple enough: you need to eliminate or reduce the thermal bridging. There are many approaches. In new construction, you can build the walls SIPs (structural insulated panels) or use advanced framing techniques which reduce the number of wall studs.

A newer approach involves applying strips of insulation over the wood studs to provide a thermal break.

In retrofit scenarios, sheets of polyisocyanurate foam, high-density rock wool, or Larsen trusses can be applied on the exterior side of the wall sheathing as a thermal break.

If your home has aluminum-framed windows, the addition of insulated shades can help reduce the severity of the thermal bridging through the aluminum during the summer.

Similarly, you can reduce the effect of thermal bridging through a metal-framed door if you install a storm door.

Thermal bridging is one of the hidden heat-loss paths that a home energy audit can uncover. Keep it in mind when thinking about the heat loss (and the heating bills) at your home.


Erik North, the owner of Free Energy Maine, is an energy auditor and home performance specialist in Westbrook, Maine. He is also the author of the Energy Auditing Blog.


  1. user-1140531 | | #1

    Objectivity of R-Value

    Referencing your article above, you said this:

    “In theory (quite literally — a material's theoretic R-value is tested under lab conditions), R-value is supposed to be an objective comparison of the thermal resistance of different materials.

    Unfortunately, this just isn’t so.”

    And in your linked article, “What is R-Value (And Why All R’s are not equal)?” you have a section titled:

    “Why is the Advertised R-value Wrong?”

    In that section, you detail ways in which insulation performance can be compromised by thermal bridging, wind washing, etc. But is the advertised R-value really wrong? I don’t believe that it is.

    I would say that R-value is indeed an objective comparison of the thermal resistance of different insulations. It is just that the total wall assembly R-value will not match the R-value of the insulation due to the compromising effect of thermal bridging and other factors. The total wall assembly will also have materials that will add R-value.

    Therefore, the advertised R-value of the insulation is one thing, and the total R-value of the wall is another thing. Just because they are different, I don’t see how the latter can be said to invalidate the former.

    Or more importantly, I do believe it is fair for you to imply (“Why is the Advertised R-value Wrong?”) that R-value of the insulation is being misrepresented because the total R-value of the wall is less than the R-value of the insulation due to thermal bridging.

    That is like saying that if you build an insulated dog house with R-10 insulation, the insulation is not delivering its rated performance because the door is open all the time.

  2. adam5532 | | #2

    The Magical Thermal Break
    There seems to be a lot of discussion of "thermal break" like it has some magical properties (or maybe I'm not understanding something). For example, when you say "A newer approach involves applying strips of insulation over the wood studs to provide a thermal break" it makes it sound like adding a thin layer of some material with good insulating properties between the stud and the drywall is going to somehow "break" the heat flow. While I understand that putting a thin layer of insulation or plastic over a wood or metal stud is going to make it feel a lot warmer (and help prevent condensation on the surface), due to its lower specific heat, isn't the calculation of heat loss through any particular slice of the wall still the sum of the R-values of each layer of material?

    Some other examples: If you build a double-stud wall (with no space in between), should it matter if you align or stagger the studs? If they are aligned, you have a continuous "thermal bridge" through the wall. If you offset them, you "break" the thermal bridge. But the heat loss of the wall assembly I believe would be the same, because instead of one 1-1/2" band of high heat loss per bay, you split it into two 1-1/2" bands of moderate heat loss ( Likewise, looking at the Mooney System, it appears like the same situation. If you take the horizontal battens and turn them vertical over the face of the studs (to "unbreak" the thermal bridge), and keep the same amount of wood inside the wall, aren't you just rearranging the areas of heat loss (now it is worse over the studs, but better between them) without changing the heat loss of the complete wall assembly? Similarly, Aerogel Thermablok strips claim an improvement of "up to 40%" (which implies 40%, but could be 1%, because it is between zero and 40!), because it breaks the "thermal bridge". But aren't the 1/4" thick strips doing nothing more than adding about R-2.5 over the edges of the studs? Which is an improvement of about 30% in just the area of the studs (R-8.5 5-1/2" stud, drywall, sheathing, siding + R-2.5 strip), which actually is significant (in that area, but probably not considering the entire wall), but only I believe because the low-loss material is able to measurably increases the total R-value of the cross section without adding much thickness, not because of some special "break" that occurs between high and low loss materials. And of course you can get another 1/4" of cellulose in the entire wall, which has nothing to do with any thermal bridge but just that you've made the wall thicker.

    Anyway, if I'm missing something, I'd be glad to hear it. But in all the articles I've read about thermal bridging, I've never seen any technical explanation on how you could possibly affect the R-value of a complete wall assembly by moving wood around inside of it, or how this "break" is nothing more than just trying to replace enough of the high heat loss material in a cross-section of wall with some low-loss material to make a dent in the overall loss of the assembly, or in some cases just adding some more insulation on the entire wall, so that the content of wood is proportionally reduced.

  3. GBA Editor
    Martin Holladay | | #3

    Response to Adam Liberman
    You're right that "the calculation of heat loss through any particular slice of the wall still the sum of the R-values of each layer of material" However, when you have a thermal bridge, you have to break down the wall assembly into different sections -- ascribing one R-value to the insulated area between the studs, and a different R-value to the area of the thermal bridges.

    Perhaps 75% of a wall consists of a sandwich of materials with insulation in the middle of the sandwich, while 25% of the wall consists of a sandwich of materials with studs and headers in the middle of the sandwich. These two sections obviously have different R-values.

    Perhaps it's easier to imagine if we consider the extreme case -- a stud wall with steel studs. If the steel studs are uninterrupted, and the studs touch the wall sheathing on the exterior and the gypsum wallboard on the interior, the studs will create cold stripes on the drywall during the winter. These stripes will be easy to see with a thermal camera. The stripes not only degrade the thermal performance of the wall; they attract moisture and create an environment for mold.

    Interrupting the thermal bridge with (for example) a continuous layer of 1-inch-thick polyiso changes everything. The cold stripe is softened -- it almost disappears from the thermal image. And the thermal performance of the wall is greatly improved.

    A so-called "Mooney wall" that is 7 1/4 inches thick performs better than a wall with 2x8 studs, because the area with continuous solid lumber is no longer 1 1/2 inch x 8 feet every 16 inches, but is instead something like 1 1/2 inch x 1 foot every 16 inches. The "cold spots" are much smaller.

  4. user-943732 | | #4

    reply to Adam
    (Note - I didn't see Martin's response posted)

    You should really just try out some basic calculations and you might see what you are missing. You would be correct if heat loss were proportional to R value, but it's proportional to 1/R and therefore the change in heat loss from adding R-1 is different depending on the R value you are adding it to.

    In your simple example of the double stud wall, let's assume the studs are R-1 per inch and the insulation is R-3. If you have the two studs offset then you have 2 sections each with R-14 (3.5+10.5) and so the UA through those two sections is 2/14 = 0.14. But if instead the studs are aligned you have one section with R-7 and one with R-21, yielding a heat loss of 1/7 + 1/21 = .0.19, which is 33% more heat loss. The average R values are the same but the heat loss is quite different. This is basic heat loss analysis. .

  5. jinmtvt | | #5

    I always wondered about
    I always wondered about omnidirectional conduction transfer.
    How is it accounted for in heat loss and R value works ??

    Surely a wood stud in a classical 2X6 with fiber batts , radiate/conduc in horizontal scheme as well as in perpendicular to wall, and the FB temperature has the same effect on the sides of the studs.

    It seems ( to me ) like if we are treating complexe wall aseemblies on a 2d scale.

    Am i wrong ? Is the exchange interface/plane the driving component ?

  6. GBA Editor
    Martin Holladay | | #6

    Response to Jin Kazama
    The short answer to your question is that when an insulation salesman brags about an R-19 wall, he is using a one-dimensional model. When we discuss the problem of thermal bridging, we are introducing a rudimentary two-dimensional model. A sophisticated energy modeling program like WUFI attempts to build a three-dimensional model.

    No model yet developed can account for all the different types of heat flow in an actual wall -- because even if it attempted to do so, there would be no way to gather the necessary inputs.

  7. adam5532 | | #7

    Response to Martin Holladay
    I am considering the area of the different sections. If we consider a double stud wall assembly that has a span of just 3" (the edge of one pair of studs, and the adjacent area): Lets say for simplicity the area of the studs is R10, and the insulated section R30. The average for the wall is R20. Now, if I offset the studs, so that the entire wall is half stud and half insulation, both areas are R20, and the average is still R20. Of course we've eliminated cold bands and helped reduce moisture problems, but isn't the heat loss the same? The study I cited (( seemed to measure no significant difference by offsetting the studs.

  8. GBA Editor
    Martin Holladay | | #8

    Response to Adam Liberman
    I see no surprises in the document you linked to. If you have a 12-inch thick double-stud wall (with two 2x4 walls separated by 5 inches of insulation), it doesn't surprise me that it hardly makes much of a difference whether the studs are aligned or offset. (After all, there are 5 inches of insulation between the two framed walls.)

    If the studs are touching -- as they would be in a 7-inch thick double-stud wall with aligned studs -- there is a difference in performance compared to a 7-inch thick double-stud wall with offset studs. Again, this is predictable. The difference isn't huge -- it's only a difference of R-1 -- but there is a difference in thermal performance.

  9. Erik North | | #9

    Ron (since it seems the other folks were addressed)

    We exchanged a few emails last year circling around the same topic. I think it stems from our backgrounds; yours as a knowledgeable high performance builder and mine as an auditor working on retrofits. Far, far too often I have conversations like, “well, I squished two 6 inch batts in this addition wall, so I have R-40, right?”

    Yes, a batt has an R-value defined in some ASTM set conditions. But those conditions hardly equate most real world states and there are so, so, so many ways the effective R-value is compromised. Moisture, air flow, installation errors, and, yes, thermal bridging.

    I don’t think the dog house analogy is very apt because an open door wouldn’t be a natural state of a residential house. But thermal bridging is a natural state of most wall assemblies. Suppose the best player on your favorite sports team tears his Achilles tendon. On paper, the team might’ve been a winner but in the real world they’re not because of the player injury.

    Thermal bridging is much the same and you should account for it. People don’t live in ASTM tested theoretical houses.

    Check out the photo on this post.

    The homeowners were so convinced that the attic was not an issue, they insisted I didn’t need to check it out.

    They had installed doubled crossed 9” fiberglass batts laid over gigantic, gaping gaps in the house’s timber framing. Was it delivering the lab tested R-60+ in heat retention? No, because it wasn’t stopping air flow.

    Don’t mean to come off combative; R-value is important. But a home is more than a layer of lab-tested insulation. And while the R-value may be objectively correct in a lab, it’s just one data point once outside of it.

  10. user-943732 | | #10

    reply to Adam
    Adam- Please read my post -- you can't average R values -- it gives you the wrong answer. You need to work with UA -- not R. The average of reciprocals is not the reciprocal of averages. .

  11. Expert Member
    Dana Dorsett | | #11

    Overstatment of bridging severity
    "The fiberglass has an R-value of around 3.5 per inch, and the stud is around 0.75 per inch."

    OK, there is fiberglass with that type of R/inch, but most species used in New England for framing run about R1.2/inch. Douglas Fir (commonly used in the Northwest) runs about R1.0/inch. I don't know of a commonly used framing species anywhere near as low as R0.75/inch.

    Which are all pretty bad, but still a heluva lot better than R0.75/inch. Sometimes facts actually matter (at least to some of us :-) ). The difference in U-factor between R1.2 /inch timber and the (non-existent) R0.75/inch timber sited results in a 60% overstatement of the actual thermal bridging.

    Reference: (See table 5.)

    Then again crummy R19 batts stuffed into an 5.5" nominal 2x6 cavity perform at only R18 according to the manufacturers' compression charts, but at least that's only a ~5% understatement of the heat loss at center cavity.

    But that doesn't change the recommended countermeasures.

  12. user-1140531 | | #12

    Reply to Erik North

    I completely agree that the real world conditions of installation, thermal bridging, etc., are entirely relevant to the performance of the insulation system in a building. I absolutely understand everything you say about that. So I agree that it is true that the advertised R-value of insulation is different than the R-value of the insulation layer in system as executed in the real world.

    However, you are not just saying that the two R-value measures are different. Instead, you are saying that the advertised R-value is wrong, and the real world executed system R-value is right.

    However, the R-value of the insulation in those various insulation systems is precisely the same as the advertised R-value of the insulation. What is different about the systems is that that insulation layer of the system will have less R-value per inch than the advertised R-value of the insulation because the system interrupts the continuity of the insulation layer.

    In your other blog piece here:
    You begin a section with the title: “Why is the Advertised R-value Wrong?” And then you go on to explain that advertised R-value is wrong because it does not account for real world compromises such as bad installation, thermal bridging, etc.

    But the R-value advertised is not claimed to represent the insulation installed wrong, or affected by thermal bridging. So how can you say the advertised R-value is wrong because it does not reflect those real world compromises? The advertised R-value is different than the R-value in real world application, but that does not mean that the advertised R-value is wrong. The advertised R-value is just a measure taken in conditions different than real world installations.

    Manufacturers could not advertise the insulation R-value if it were to include real world application compromises, because the compromises vary from one application to another. So the advertised R-value is not wrong. It is simply a measure of performance taken ahead of the real world application compromises, in order to compare insulation performance alone on an apples-to-apples basis.

    The dog house analogy is apt because it would be unfair to blame the insulation performance on the open door. And to do so would be analogous to blaming insulation performance in a real house on missing or poorly installed insulation.

    Perhaps this issue could be resolved by changing your question, “Why is the Advertised R-value Wrong?” to “Why is the advertised R-value not delivered in some real world applications?”

  13. GBA Editor
    Martin Holladay | | #13

    Response to Dana Dorsett
    You wrote, "most species [of softwood lumber] used in New England for framing run about R-1.2/inch."

    You're right, of course, as Erik North just acknowledged. Erik just sent me an e-mail asking me to correct the R-value for studs in the blog, and I have made the correction.

    Thanks, Dana, for pointing out the inadvertent error.

  14. adam5532 | | #14

    Reply to Michael Blasnik
    Aha, I did the calculations and I stand corrected! Thank you!

  15. badgerboilerMN | | #15

    5" of 2# foam can't stop
    5" of 2# foam can't stop thermal bridging here in Minneapolis, as evidenced by picture of my recently remodeled 1921 farm house. Had I only followed Dana a year sooner...

    Sorry, great picture but maximum file size will not handle a postage stamp. Suffice it to say, striping is evident.

  16. user-731642 | | #16

    Structural Issues
    Great comments on thermal bridging, but did anyone notice that the cross-bracing does not appear to extend to the top of the wall?

  17. runner9 | | #17

    Raising attic floor for more insulation
    I'm planning to run 2x10 rafters perpendicular to the current attic floor so I can add insulation and still use it for storage. Is it worth it to put 2 inch rigid foam between the new 2x10 and the current attic rafters at the points where they cross to try to decrease any thermal bridging? I assume future weight would only compress the rigid foam a little, it's only for storage anyway.

  18. GBA Editor
    Martin Holladay | | #18

    Response to Jeremy M
    I assume that you are talking about joists (floor framing), not rafters.

    If you install the new joists perpendicular to the existing joists, you'll eliminate almost all of the thermal bridging. There is no need to install little blocks of rigid foam at the intersections.

Log in or create an account to post a comment.



Recent Questions and Replies

  • |
  • |
  • |
  • |