What counts as an explanation?

Imagine standing beside a patch of mussels as the tide falls. A moment ago, water covered the shells. Now the animals are exposed to air. Someone says, “They close because they don't want to dry out.” The sentence sounds sensible. It also bundles an observation, a claim about a benefit, and a claim about an animal's experience into nine words. You could agree with it and still know almost nothing about how the animal works.

We will use this shore as a starting point for a different habit: keep the puzzle interesting while making the explanation more exact. A mussel need not look busy to present demanding biological questions. Its body has to obtain materials, support living cells, withstand changing conditions, and pass through a life cycle. Understanding those tasks begins with deciding what kind of answer a question requires.

Begin with what was actually observed

A description records a feature or event. “The gap between the shell edges became smaller” describes something a camera might record. “The animal became frightened” interprets the change through a proposed internal state. That interpretation might eventually be investigated, but it is additional to the visible movement. Good observation gives other people enough information to recognize the event without first accepting your explanation.

The Monterey Bay Aquarium identifies the California mussel as Mytilus californianus and describes its contrasting exposure and feeding conditions: submerged animals can open their shells while beating cilia draw water carrying food particles. This gives us an identified animal and a documented starting point. It does not provide a complete account of every closure you might witness. Monterey Bay Aquarium: California mussel

Suppose an observer writes, “Eight of ten visible shells had a narrower gap two minutes after the water receded.” The numbers here are invented, but the form of the record matters. It states the sample, change, and interval. It leaves room for the other two animals and for uncertainty about how well the gaps could be seen. “Mussels always close at low tide” discards those useful details.

An observation also needs a setting. A shell hidden behind another shell is not necessarily closed. A still photograph cannot establish how quickly a gap changed. A view of ten animals in one patch cannot describe all ages, shores, or weather conditions. These are limits of the information collected, not reasons to abandon observation.

Ask what makes the change happen

A mechanism explains how interacting parts and processes produce an event. For shell closure, a mechanistic investigation would have to connect a relevant input to the activity that changes the shell's position. Naming “the nervous system” is a possible starting location, not a finished causal chain. Which input was detected? What changed in the animal? How did that change produce force or motion?

Think of a diagram with four boxes: environmental change, detection, internal response, visible movement. Each arrow makes a claim. If all you measured was the first and last box, the two middle boxes remain proposed steps. Filling them with impressive vocabulary does not turn them into observations. The diagram becomes more useful when each box has a possible measurement attached.

In an imagined study, the relevant measurements might include local conditions, the timing of a response, and the activity of a specified tissue. We are outlining questions, not proposing that a reader perform invasive experiments. The point is to see what evidence a complete mechanism would demand. A continuous recording of shell movement would improve the last box; it would not by itself identify the detection process.

Mechanisms can also have several inputs. An internal state could alter the response to the same external change. A claim such as “this cue always produces closure” predicts more than “this cue changes the probability or speed of closure under these conditions.” Choosing between those claims changes what results would count against the explanation.

Separate operation from development and history

How something works now is one question. How an individual acquired its structures and responses is another. Development concerns the processes through which an organism changes during its life. A useful developmental account might compare stages, trace the formation of a structure, or examine how an earlier experience affects a later response. An adult photograph cannot supply that sequence.

Evolutionary history concerns changes across generations and relationships among lineages. It asks how the features seen in living organisms came to have their present distributions. A similarity between two animals may suggest a question about shared ancestry; resemblance alone does not establish the answer. A difference between young and adult individuals is not, simply by being a difference through time, evolution within that individual.

Niko Tinbergen's 1963 discussion of animal behavior distinguished causation, development, survival value, and evolution while arguing that they should be connected. The distinction prevents a useful answer to one problem from being mistaken for an answer to all of them. Tinbergen: On aims and methods of Ethology, page 411

Return to the shell gap. A record of how a young animal's response changes as it grows could answer a developmental question. A comparison of responses in related species could contribute to a historical investigation. Neither automatically reveals the immediate process that closed this shell at this moment. Keep the time scale in the question visible.

A benefit is a claim that can be examined

“Closure reduces water loss” proposes a consequence. Unlike a claim about wanting, it suggests quantities that could be compared. What water is lost, over what interval, under which conditions? Relative to what alternative? Even before knowing the answer, those questions make the proposal more informative than “the shell protects the animal.” Protection could mean several very different things.

A current benefit and an evolutionary origin also differ. Showing that a feature has a useful effect today does not establish the historical process that produced it. A feature can have several effects, and its present use need not have been its original use. Bateson and Laland emphasize this distinction in their reassessment of Tinbergen's framework. Bateson and Laland: current utility and evolutionary explanation

Consider an invented comparison in which one shell arrangement loses less water than another but also admits fewer food particles during a particular interval. Calling the first arrangement “better” would conceal a decision about which consequence matters. A useful account would report both outcomes and the conditions under which each becomes important. Chapter 6 will develop this problem as a tradeoff across a life cycle.

It is possible to be precise without pretending that every effect has already been measured. “Reduced water loss is a hypothesis consistent with this pattern” identifies the status of a claim. “The animal evolved the behavior to avoid drying” adds a historical conclusion for which the pattern alone supplies no test.

Natural selection requires a population account

Natural selection connects variation, inheritance, and differences in reproductive contribution. If a heritable difference systematically affects how much individuals contribute to later generations, the distribution of that difference can change. The UC Museum of Paleontology introduces these ingredients together; usefulness without inheritance and reproductive consequences is not the whole account. UC Berkeley: natural selection

Here is a deliberately simplified numerical model. Begin with fifty individuals of type A and fifty of type B. Assume each A contributes, on average, three surviving descendants to the next counted generation, while each B contributes one. Further assume descendants retain their parent's type, and ignore other population processes. The next count contains 150 A and 50 B: A has increased from half to three-quarters of the population.

Nothing in that arithmetic says an individual decided to become A. It also does not say why A contributed more descendants. To connect the model to an actual animal, you would need evidence about the relevant difference, its transmission, and the circumstances affecting reproduction. If the apparent types instead reflect a temporary condition that descendants do not inherit, the model's assumptions have failed.

Nor does three-quarters mean “the most advanced animal.” It is a proportion in a particular population under specified conditions. Change the conditions or the reproductive outcomes and the next calculation can change. A model becomes a biological explanation only through its justified connection to the organism and environment being studied.

Make competing explanations disagree

An explanation becomes easier to assess when you identify another explanation that predicts something different. Imagine a fictional recording in which shell gaps narrow after the tide falls. One proposal is that exposure to air supplies a relevant cue. Another is that a coincident change in local water movement matters. The recording may be compatible with both. Repeating “after low tide” does not separate them.

A conceptual comparison would vary the two candidate conditions independently, if that could be done appropriately in a research setting. It would ask about exposure with one flow condition, exposure with another, and corresponding submerged conditions. The value of this design lies in the contrast it creates. It gives the hypotheses opportunities to diverge instead of allowing either to explain every possible result afterward.

Other differences must still be considered. If all exposed animals are small and all submerged animals are large, size is entangled with the main comparison. If every observation in one condition occurs on a single unusually warm day, the day is also a candidate influence. Replication means more than producing many numbers from the same unchanging setup.

You can apply this reasoning to a paper without running an experiment yourself. Write the proposed cause, the comparison, and the predicted result before reading the conclusion. Then ask whether the actual measurements distinguish the explanation from its nearest alternative. This turns a scientific claim into something you can inspect rather than simply accept or dismiss.

Count the right unit

Suppose the fictional observer takes one photograph every minute for an hour. There are sixty photographs of the same ten animals. Has the study now observed six hundred independent animals? No. Repeated measurements help describe those animals through time, but they do not create additional individuals. The difference matters when judging how broadly the result can travel.

Now suppose the ten animals all occupy one shaded crevice. Their shared location could explain part of their shared behavior. Adding ten more photographs from the same crevice improves the time record but does little to compare exposed and sheltered locations. A study's unit of measurement, the unit receiving a different condition, and the population it hopes to describe need not be the same.

An average can conceal a second problem. If five animals respond rapidly and five do not respond within the observation period, an average response time may depend on how the nonresponses were treated. Assigning them the final observation time would mean something different from measuring an actual closure then. Ask how the numbers were constructed before interpreting a neat summary.

This is why a good animal account retains variation. Variation may reflect measurement limits, meaningful differences among animals, or changing conditions. It is part of the phenomenon to explain. Removing it too early can make a weak mechanism look universal and can hide precisely the comparison that would teach us more.

Follow an explanation across levels

An organism's behavior happens through a body, and a body is made of living cells. Yet an answer at one level does not automatically replace an answer at another. Describing a cell's response could explain one link in a larger process. It would still leave questions about how many cells act together, how the resulting tissue interacts with other structures, and what consequence follows for the animal.

Imagine explaining the changing shell gap using only the phrase “molecules interact.” The phrase is compatible with the event, but it cannot distinguish closure from opening. At the other extreme, “the mussel survives the tide” names an outcome while skipping the working processes. A useful explanation connects levels at the places where the connection matters to the question.

One way to test your account is to remove an arrow from its diagram. If the final claim still appears to follow just as well, perhaps that arrow was decoration. If the removal leaves a clear missing step, you have identified where evidence is needed. The aim is not to include every molecule. It is to show enough of the chain that a reader can understand what would change if a key step changed.

The next chapters will supply the physical and biological tools for those connections: cells and surfaces, size and distance, energy and exchange, regulation, and life cycles. Keep this opening distinction with you. A body becomes more interesting as its explanation becomes more demanding, because each resolved question reveals another process that can actually be investigated.

Application

Rewrite the shoreline claim

Begin with: “Mussels close because they know the tide is going out and evolved to protect themselves.” Write four separate sentences: one observable description, one mechanistic question, one proposed present consequence, and one historical question. Do not turn the proposed answers into facts. You may use the documented feeding context above, but the numerical observations in this chapter remain fictional.

Audit an invented study

A report shows sixty photographs of the same ten mussels in one crevice. Eight have narrower shell gaps after exposure begins. The author writes, “Six hundred observations prove all mussels respond to air rather than water movement.” Identify three problems with the conclusion. Propose one comparison that would help separate the candidate explanations, without giving instructions for disturbing living animals.

Model interpretation

A suitable description is: “In the recording, the visible shell gap narrowed in eight of ten animals after the water receded.” A mechanistic question is: “Which detected change and internal process produced that movement?” A proposed consequence is: “Does the changed gap alter water loss under these conditions?” A historical question is: “What evidence connects variation in this response, inheritance, and reproductive consequences in the lineage?” These sentences divide an attractive story into investigable claims.

The study repeatedly measured ten animals; it did not observe six hundred independent animals. One crevice cannot establish a response for every mussel or setting. Exposure and changed water movement coincide, so the recording does not distinguish them. A suitable research design would compare the candidate conditions separately while accounting for shared location and other relevant differences. More photographs alone would not repair those gaps.

For the numerical selection model, 150 divided by 200 is 0.75. The calculation follows the stated transmission and reproductive assumptions. It demonstrates what those assumptions imply; it supplies no evidence that a real shell response follows this inheritance pattern.