How to Run the Field Tests That Classify Soil

Why this matters

Every one of these tests is run from outside the cut, and they are run by the competent person, meaning under 29 CFR 1926.650(b) someone who can identify the hazard and who has authorisation to stop the work over it. A test run by anyone else produces a reading, not a classification. Nobody climbs into an unprotected excavation to reach a face or scoop a sample; the material comes out with the bucket, off the spoil, or off the face using a tool from the surface, and the face itself is read from the top. That is the whole point: the reason you are testing is that you do not yet know whether the wall will stand.

The tests take a couple of minutes and cost nothing. What separates a classification from a gesture is where the sample came from, how fresh it was, whether you ran both kinds of analysis Appendix A requires, and what you do when two tests disagree. This card is those four things; the classes themselves and the conditions that force a downgrade are owned by a sibling card.

Before any sample is taken

Locate underground installations first. 29 CFR 1926.651(b) requires the estimated location of utility installations to be determined before opening an excavation, and utilities to be protected, supported or removed while the excavation is open. No sample is worth striking a service.

Handle the material with the exposure in mind. Soil in an occupied area can carry sewage, fuel or fertiliser contamination, so wear gloves for hand tests and do not eat, drink or smoke before washing. Crumbling a dry sample raises dust that can carry respirable crystalline silica where the material is dry and granular, so crumble it downwind at arm's length rather than at your face. The construction silica standard at 29 CFR 1926.1153 is written around the equipment tasks that generate the real airborne loads rather than around handling a clump, but the control for your own breathing zone is the same one.

Stay off the spoil toe. A fresh pile ravels at the bottom and it sits at the edge of your own cut. Take the clump from the side you can reach with both feet on undisturbed ground.

Step 1: get the sample from the right place, at the right time

Appendix A's thumb penetration test says to use an undisturbed sample, such as a large clump of spoil, as soon as practicable after excavation, to minimise the effects of drying. That clause governs the whole exercise, because drying is not neutral. A cohesive sample loses moisture, gains apparent hardness at the surface, and opens fissures, and both of those errors run the same way: they make the soil look better than it is.

So the sample is a clump that came out in the last few minutes, taken whole rather than broken apart, and tested before it sits in the sun. A sample pulled off a pile cast yesterday tests the weather, not the ground. Take samples from each segment of the run and from each layer visible in the face, because one sample classifies one condition.

Step 2: run the visual analysis, and run it first

Appendix A requires at least one visual and at least one manual analysis. Run the visual first, because it can end the classification before you touch a sample: a seepage face or a road with truck traffic settles the class regardless of how strong the clay is.

Working from the surface, look for five things.

Utilities and previous disturbance. Patched pavement, a strip of different backfill in the face, a marker post, a change in colour or texture. Previously disturbed soil is Type B at best and Type C where the backfill is loose or granular.

Fissures. Cracks, seams and separations in the exposed face. Look again later in the shift; a clay face opens fissures as it dries, so the morning inspection does not cover the afternoon.

Layers and their dip. Identify each layer in the opened side and estimate the direction and steepness of the dip. Layers dipping into the excavation at 4 horizontal to 1 vertical or steeper are a Type A disqualifier and can be a Type C trigger.

Water. Surface water, seepage from the sides, the water table, and anything that discharges nearby - a downspout, a sump line, an irrigation head. Water always makes the class worse and never better.

Sources of vibration. Traffic, rail, compaction, pile driving, on your site or next to it. Note what is running now and what runs later in the day.

Watch the material as it comes out too: soil that stays in large clumps is cohesive, soil that breaks up and does not hold a clump is granular. That tells you which manual tests are worth running.

Step 3: run the manual tests that answer your open question

Each test answers one question. Pick the ones that close the questions the visual left open, and run at least one.

Is it cohesive or granular? Plasticity, the ribbon or thread test. Mould a moist sample into a ball and roll it into threads about 1/8 in in diameter. If a 2 in length of that thread can be held by one end without tearing, the material is cohesive. Granular material will not form the thread at all.

Is it cohesive or granular, when the sample is dry? Dry strength. A dry sample that crumbles on its own or under moderate pressure into individual grains or powder is granular. One that falls into clumps which break into smaller clumps only with difficulty may be clay in some combination with gravel, sand or silt. One that breaks into clumps that will not break down further, with no visual indication of fissuring, may be considered unfissured.

Is it fissured? The drying test. Dry a sample roughly an inch thick and six inches across until it is thoroughly dry. Cracks developing as it dries indicate significant fissures. This is the test that catches fissuring a face has not yet shown you.

How strong is it? Thumb penetration. On a fresh undisturbed clump: material at about 1.5 tsf can be readily indented by the thumb but penetrated only with very great effort; material at about 0.5 tsf can be easily penetrated several inches by the thumb and moulded by light finger pressure. It is a coarse test and honest about being one - it puts a sample in a band, not on a number.

How strong is it, with an instrument? A pocket penetrometer or a hand shear vane. These read strength directly and they are the only tests here that produce a figure.

Step 4: know what your strength readings actually are before you use them

A penetrometer reading is a number, and a number invites arithmetic it does not support. Two labels go on it first.

Its basis. The reading is in tons per square foot on that one sample at that one spot. It is not a percentage of anything, so it does not scale, and the spread across samples is a measured spread of the ground, not an instrument specification.

Its character. Sample-to-sample variation is an independent spread reflecting real variability in the soil. Operator technique - how fast the plunger is pushed - is a systematic offset that does not cancel between two people, so one person takes the whole set, and a set taken by two people is two sets.

That has one consequence: do not average. A mean lets a strong sample carry a weak one, and the weak one is the material that will be in the wall somewhere. Take the lowest defensible reading as the governing value and state it as a bound with one inequality - "strength is not demonstrated above 1.2 tsf" - rather than as a mean with a plus-or-minus, because you are not estimating the soil's central value, you are establishing what it is entitled to be treated as.

Step 5: when tests disagree, the lower class governs

One sentence: where two analyses point to different classes, classify at the lower one. No tie-break, no best-two-of-three, no judgement that splits the difference. The reason is asymmetric consequence - one class low costs width and time, one class high puts a face up steeper than the ground will hold, and the warning before that fails is measured in seconds.

The protocol run on one cut

A residential water service replacement: 6 ft deep, 3 ft wide, running from the meter box at the sidewalk to the house, crossing a concrete drive.

What the mean would have done. The three readings average to about 1.43 tsf, which also lands below the 1.5 tsf line, so on this cut the mean and the lowest agree. Do not take comfort from that. Shift one reading and the mean crosses the line while the lowest does not, and the crew that uses means will get a Type A out of a soil whose weakest sample is Type B. The rule is not "the mean happened to work here," it is "the lowest governs."

What the classification then bought. At 6 ft the cut is past the 5 ft trigger, so a protective system is required. Type B permits 1 to 1, which is 6 ft of run each side, so a sloped cut is 15 ft wide at the surface - between a house and a sidewalk that usually does not fit, which sends the job to a support or shield route. Sibling cards own that geometry.

The failure mode. The thumb test alone would have produced Type A, and a crew that ran one manual test and stopped would have had a defensible-looking classification that was one class high. The protection against that is not a better thumb, it is running a second test that produces a number and letting the lower answer win.

When to run it again

Re-run the analysis, not just the inspection, whenever any of these happens: rain or snowmelt, a new stratum in the face, the alignment crossing into different ground, equipment or traffic starting up within influence of the cut, spoil placed where it was not before, or a cut left open long enough for the face to dry. 29 CFR 1926.651(k) sets the inspection cadence - before the shift, as needed through the shift, and after every rainstorm or other hazard-increasing occurrence - and each of those is an opportunity to find that the class has moved.

Record the segment, the time, the tests run, the readings, the class, and who made the call. That record is what later shows the classification was made rather than assumed, and it is what tells the next crew which segment they are standing in.

References

• Appendix A to 29 CFR 1926 Subpart P, soil classification, including the visual and manual test methods, the requirement for at least one of each, and the as-soon-as-practicable sampling condition.
• 29 CFR 1926.651(b) for locating underground installations before opening ground, and 29 CFR 1926.651(k) for the inspection and re-inspection triggers.
• 29 CFR 1926.1153, respirable crystalline silica in construction, for the dust-generating tasks that carry a formal control obligation.
• See related: the library's cards on how soil is classified and what each class permits, on sloping and benching by soil type, and on what a protective system is required to do.