How Insulation Actually Works

Why this matters

Insulation is the one product in the trades sold by a single printed number, and that number is a laboratory result taken under conditions almost no installation reproduces. Shops that add thickness without checking the four assumptions behind the number produce upgrades that measure worse than the customer paid for, and the customer's own experience tells them so long before any energy bill does. Understanding the mechanism is what lets you predict which upgrades will land and which will not, on a house you have never seen.

The arithmetic of adding R-values in series, and the code minimums, belong to a sibling card. This one is about what is physically happening and what breaks it.

What the material is actually doing

Insulation does not block heat. Nothing blocks heat. It slows heat by attacking the three modes at once, and the largest part of that work is done by gas rather than by the material you are holding.

Still air is an excellent insulator. Air conducts at roughly 0.0146 Btu per hr per ft per F near room temperature, which works out to about R-5.7 per inch if you could hold it perfectly motionless. That is better per inch than most fibrous insulation on the shelf. The reason a batt is rated lower than the air inside it is that the fibers themselves conduct a little and the batt has to carry its own weight.

So why not just leave an air gap? Because air will not hold still. Give it more than roughly three quarters of an inch of open vertical space with a temperature difference across it and it starts a convection loop, warm air up one face and cool air down the other, which carries far more heat than conduction would have. The fibers or cells exist to stop the air from circulating, not to block heat themselves. Every insulation product on the truck is a different mechanical answer to the same problem: hold gas still.

That reframes the whole category:

• Fibrous products (glass wool, mineral wool, cellulose) trap air in the spaces between fibers and scatter radiation off the fiber surfaces. They depend entirely on the air staying put, which is why they are vulnerable to air movement through and around them.
• Closed-cell foams seal the gas inside discrete cells, so it cannot circulate at all, and many use a gas with lower conductivity than air. That gas advantage is not permanent: it diffuses out and air diffuses in over years, so the product's aged or long-term rating, not its fresh one, is the number to design with.
• Radiant barriers attack only the radiant mode, with a low-emissivity surface. They do nothing unless they face an air space, because a foil pressed flat against a surface is simply in contact with it, and contact is conduction. A dusty foil has also stopped working, since dust raises emissivity.

The four assumptions inside every printed R-value

A rated R-value assumes all four of these at once. Field underperformance is almost always one of them being false, not the material being wrong.

1. Full rated thickness, uncompressed. R scales with thickness. Compressing a product into a shallower space loses more than it gains from the higher density: a batt rated for a deeper cavity, stuffed into a shallower one, commonly lands well below its label, and the manufacturer's compression chart is the authority on how far.
2. The trapped gas is still. Air moving through the material, or across its face, or looping through a gap behind it, defeats the mechanism directly.
3. No parallel high-conductivity path. R-values in series add. Paths in parallel do not add, they compete, and the low-resistance one wins far out of proportion to its area.
4. Dry. Water conducts on the order of twenty-five times better than air. Wet fibrous insulation is not degraded insulation, it is a different and much worse material, and it also loses loft permanently.

Before you go into the space

Every one of these checks happens somewhere unpleasant, and the hazards are specific.

• Falling through a ceiling. In an attic, weight goes on framing members or laid planking, never on the ceiling membrane, and the path is planned and lit before you commit. Fall protection duties sit at 29 CFR 1910.28 for general industry work and 29 CFR 1926.501 for construction work; a field-service shop can fall under either depending on the job, so know which applies to yours before you rely on it.
• Airborne fibers and dust. Disturbing loose-fill, batts or blown material puts respirable dust in the air at head height. The control for an inhalation route is a respirator, not a glove: at minimum a NIOSH-approved filtering facepiece, used under a program meeting 29 CFR 1910.134 including the annual fit test at 1910.134(f)(2), plus eye protection under 1910.133. Gloves and sleeves address the skin irritation, which is the separate, lesser problem.
• Suspect material stops the job. Thermal system insulation on pipes, boilers and ducts installed before 1981 is presumed asbestos-containing and is not to be disturbed, cut, scraped or removed by anyone outside a qualified program: 29 CFR 1910.1001 for general industry, 29 CFR 1926.1101 for construction. Loose granular material with a pebble-like appearance in an attic is the other stop-work case; leave it, and tell the customer in writing why.
• Heat. An attic in summer reaches temperatures that put a tech down fast. Work it early, in pairs, with water staged; the sibling cards on heat illness carry the protocol.
• Anything electrical in the cavity. Junction boxes, knob-and-tube, and buried splices are live until proven otherwise. De-energize and lock out under 29 CFR 1910.333(b)(2) and prove dead with the live-dead-live sequence in NFPA 70E-2021, 120.5 before insulation is moved off or packed around any of it.

The callback: more R, colder bedrooms

A shop blows an attic up to a nominal R-40 over roughly 1,000 ft2 of ceiling. Six weeks later the customer calls back: the upstairs bedrooms are no better, one is worse, and there is a stain at the ceiling edge.

The tech who goes back does three things the installer did not: he measures depth at the eaves as well as the middle, he reads ceiling surface temperatures across the whole room rather than in the center, and he counts the penetrations.

What the surface scan showed. The center of each ceiling read within a degree of the room. A band about a foot wide around the perimeter read several degrees colder, and there were four distinct cold discs in the pattern of the recessed fixtures, plus one large cold rectangle at the attic hatch. The cold discs and rectangle are parallel paths. The cold perimeter band is the second assumption failing: outside air entering at the eave vents and washing through the first few feet of loose-fill, plus depth that tapers to almost nothing where the roof deck meets the top plate.

The parallel paths, in arithmetic. Work in conductance, which is the reciprocal of R, because conductances in parallel add and R-values do not.

• Insulated ceiling: 1,000 ft2 at R-40 gives U = 1 / 40 = 0.025, so 25 Btu per hr per F
• Attic hatch: a bare panel with its air films is near R-2, so U = 0.5. Over 4 ft2 that is 2 Btu per hr per F
• Four recessed fixtures held clear of insulation, roughly 0.6 ft2 of opening each, so 2.4 ft2 also near R-2: 1.2 Btu per hr per F

Total: 25 + 2 + 1.2 = 28.2 Btu per hr per F over 1,006.4 ft2, which is an effective U of 0.028 and an effective R of about 36.

Read what that says. The hatch and the fixtures together are 6.4 ft2, about 0.6 percent of the ceiling area, and they added 3.2 / 25 = about 13 percent to the ceiling's heat flow, pulling a nominal R-40 assembly down to roughly R-36. The customer bought R-40 and the ceiling behaves like R-36 before anyone accounts for the perimeter wash, which is not in these numbers at all.

Why the arithmetic has to be done in conductance. A tech who averages R-values by area gets a comforting wrong answer: 0.6 percent of the area at R-2 barely moves an area-weighted R-40. Averaging resistances treats the paths as though heat has to cross both, and it does not, it takes whichever one is easier. The conductance form is the one that matches what heat actually does, and it is the reason a small bypass matters so much more than it looks like it should.

What was actually wrong, in order of size. The perimeter wash and the taper at the eaves were the largest single defect and the one the customer felt in the bedroom over the eave. The fixtures and hatch were the next, and they are cheap to correct with rated covers and a gasketed insulated hatch. The blown depth in the middle of the field was fine, which is exactly why the installer's own inspection passed: he measured where the material was deepest.

The failure mode to take from this. A depth check in the middle of an attic verifies nothing about the assembly. The heat leaves at the edges and through the holes, and both of those are places a tape measure does not naturally go.

What would change the conclusion. If the surface scan had shown a uniformly cold ceiling with no perimeter band and no discs, the finding would be thickness or coverage across the whole field, and the fix would be more material. Uniform means material; local means bypass. That single distinction decides which quote you write.

Wet, compressed, and drifted

Three failures do not show as holes in a scan and get missed for years.

• Wet. A roof leak, a bath fan discharging into the attic, or condensation on a cold surface soaks the material. It reads as a diffuse cool region rather than a sharp one, and it does not recover its rating when it dries because the loft is gone. Water is also the start of a mold and structural problem that outranks the energy question.
• Compressed. Anything stored on top, anything walked on, wiring pulled tight across a batt, or a batt oversized for its cavity. The material looks present and is performing at a fraction of label.
• Drifted. A closed-cell foam whose blowing agent has partly diffused out is performing at its aged value, not the fresh one. This is normal product behaviour rather than a defect, and it is why the long-term rating is the design number.

None of the three is visible from a label, an invoice or a photograph of the installation day. All three are visible from a surface temperature scan taken at a real temperature difference.

What to measure instead of trusting the label

• Take surface temperatures at a real difference, the larger the better. On a mild day the whole assembly reads uniform and tells you nothing. Early morning in winter, or a hot afternoon in summer, is when defects show.
• Scan the perimeter and the penetrations first, not the field. That is where the conductance arithmetic says the losses are concentrated.
• Measure depth at three places minimum, including the eave and one point you have to crawl for. A single center measurement is the reading that lets a bad installation pass.
• Count and list every penetration, then put them in a conductance table beside the field area. The total tells you whether sealing the penetrations or adding depth is the better spend before you quote either.
• Verify the material is dry by touch through a glove and by weight where you can, and look up: a wet spot in insulation has a source directly above it.

References

• ASHRAE Handbook Fundamentals, thermal properties of insulating materials, air spaces and surface films
• Manufacturer product data for compression charts and aged or long-term thermal resistance
• 29 CFR 1910.134 including 1910.134(f)(2), respiratory protection for airborne fibers and dust; 29 CFR 1910.133, eye protection
• 29 CFR 1910.1001 and 29 CFR 1926.1101, asbestos in general industry and construction, including presumed asbestos-containing thermal system insulation
• See related: Insulation R-Value Math and Application Reference; How Heat Actually Moves and Why It Matters on a Call